Eye spy with my little AI: An introductory look at ChatGPT and Google Gemini for ophthalmologists

Dear Editor, The insurmountable rise of artificial intelligence continues to impact ophthalmologists around the world. The scope of AI is not just limited to the diagnosis and treatment of diseases but is now explored in a screening where disease likelihood and progression can be predicted with little to no clinical findings.[1] Conversational AI language models, such as ChatGPT and Google Gemini, are trained in an ocean of textual data making them capable of understanding natural language input and providing answers accordingly. They are also used in the field of academics, scientific literature and research.[2] We looked into both chatbots as a resource for ophthalmologists and came to the following conclusions. Gemini provides updated information because of its association with Google’s extensive search network unlike ChatGPT (Version 4) whose information is dated only until April 2023.[3] Although both chatbots show similarities in terms of natural language processing abilities and human-like responses, Gemini takes it one step forward by producing additional content such as images and reference links with the answers. ChatGPT’s intuitive data search and ability to generate meaningful sentences have been utilized in many articles in peer-reviewed medical journals where it has been credited authorship.[4] However, both these models come with limitations like the inability to understand wrong inputs, inadequacies in inferential capabilities and providing inaccurate responses that sound plausible, called ‘Hallucinations’.[5,6] For the purposes of this letter, we asked both chatbots a similar question and found that Gemini provides concise and articulated outputs when compared to ChatGPT [Figure 1]. Both AI tools assist in academics and research, but we have yet to fully explore their capabilities. Google Gemini is vastly more advanced than its predecessors in terms of language processing and response generation. It proves a worthy competition for ChatGPT in aspects like data, voice and image processing as well as providing updated information for the same. But bearing in mind their limitations, we discourage the use of these models in academic literature without a comprehensive fact check, due to the inevitable plagiarism and biases they display.[7]Figure 1: Table depicting answers provided to the same prompt by Chat GPT (Version 3.5) and Google Gemini (2024)Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.