Recent advances in Artificial Intelligence (AI) have enabled better triaging of ophthalmic symptoms both quickly and accurately. Tsui et al. recently researched the use of an online AI chatbot, ChatGPT, for ophthalmic triage in their study. The authors designed ten different scenarios for common ophthalmic symptoms including retinal detachment, age-related macular degeneration, and ocular surface disease. Though the results of this study were encouraging, with ChatGPT providing suitable and precise responses for eight of the ten scenarios, it failed to provide satisfactory answers for the remaining two.
In response to these findings, OpenAI has released GPT-4: designed with a broader knowledge base, image analysis capabilities, and superior problem-solving ability. To assess GPT-4’s capabilities, the team revisited the two scenarios that ChatGPT struggled with, providing precise and suitable replies, as can be seen in Table 1. For example, when prompted “I have macular degeneration and things look wavy today”, GPT-4 appropriately recommended seeking immediate medical advice – advice which ChatGPT would not provide.
Future uses of AI has been identified in many areas of medical care, such as ChatGPT being used in medical emergency situations during long-duration space flights [3]. Additionally GPT-4’s image analysis capabilities have prompted greater interest in harnessing AI for ophthalmic image triage.
However, it is important to note that caution must be taken when using the technology in a clinical setting due to certain limitations [4]. Nevertheless, the groundbreaking research conducted by Tsui et al. should be applauded, and further understanding from the authors as to the differences between ChatGPT and GPT-4 would be of great value.
