Answers are grounded in this paper and cite their sources.
We commend Seth et al's article highlighting the use of GPT-4 (OpenAI, San Francisco, CA) and assessing the chatbot's efficacy at answering frequently asked questions (FAQs) pertaining to breast augmentation.1 The authors used an expert panel of plastic and reconstructive surgeons along with a literature search via PubMed and Cochrane to qualitatively evaluate the outputs to the FAQs they prompted. GPT-4 was able to provide accurate information but lacked personalization in its answers, which is a major limitation of current artificial intelligence systems. Undergoing a procedure can be anxiety inducing and patients will naturally have questions they want addressed.2 It is important to ensure patients are appropriate candidates for surgery and that their concerns are heard in a setting that is free from judgment but also underscoring the risks and benefits of the operation. Among their range of questions, the authors prompted the chatbot regarding the risks and complications that could result from breast augmentation, which returned a succinct output containing limited medical jargon. GPT-4's output covered a list of complications while discussing the need for reoperation or revision surgery, temporary or permanent changes, malposition, and breast implant–associated anaplastic large cell lymphoma (BIA-ALCL). With respect to BIA-ALCL, it gave an accurate range of 1 in 4000 to 1 in 30,000, with textured implants being associated with increased risk.3‐5 The chatbot’s emphasis of consulting a board-certified plastic surgeon in its answers is promising and agrees with our belief that these tools, particularly in their infancy, should be used to assist surgeons. The chatbot is also capable of highlighting its own limitations with respect to its knowledge (eg, referencing its 2021 cutoff training date) and uses statements such as “I am not a doctor …”; it also encourages the user to seek the advice of a medical professional.
No takes yet. Share an insight, caveat, or question.
Najafali et al. (2023) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: