Department of Urology, Chengalpattu Medical College, Chengalpattu, India.
RediMinds, Inc., Southfield, Michigan, USA.
Can J Urol. 2023 Aug;30(4):11588-11598.
Artificial generative intelligence (AGI) and large language models (LLMs) have gained significant attention in healthcare and hold enormous promise for transforming every aspect of our life and urology is no exception.
We conducted a comprehensive literature search of electronic databases and included articles discussing AGI and LLMs in healthcare. Additionally, we have incorporated our experiences interacting with the ChatGPT and GPT-4 in different situations with real case reports and case constructs.
Our review highlights the potential applications and likely impact of these technologies in urology, for differential diagnosis, prioritizing treatment options, and facilitating research, surgeon, and patient education. At their current developmental stage, we have recognized the need for concurrent validation and continuous human interaction necessary to induce inverse reinforced learning with human feedback to mature them to authenticity. We need to consciously adjust to the hallucinations and guard patients' confidentiality before their extensive implementations in clinical practice. We propose possible remedies for these shortcomings and emphasize the critical role of human interaction in their evolution.
The integration of these tools has the potential to revolutionize urology, but it also presents several challenges needing attention. To harness the full potential of these models, urologists must consistently engage in training these tools with their clinical sense and experience. We urge the urology community to actively participate in AGI and LLM development to address potential challenges. These models could help us in unleashing our full potential and help us achieve a better work-life balance.
人工智能（AGI）和大型语言模型（LLMs）在医疗保健领域引起了广泛关注，它们有望改变我们生活的方方面面，泌尿外科也不例外。
我们对电子数据库进行了全面的文献检索，纳入了讨论医疗保健中的 AGI 和 LLM 的文章。此外，我们还结合了我们在不同情况下与 ChatGPT 和 GPT-4 交互的经验，包括真实病例报告和病例构建。
我们的综述强调了这些技术在泌尿外科中的潜在应用和可能的影响，可用于鉴别诊断、优先选择治疗方案以及促进研究、外科医生和患者教育。在它们目前的发展阶段，我们已经认识到需要进行并行验证和持续的人机交互，以诱导带有人类反馈的反向强化学习，使它们成熟并具有真实性。在广泛应用于临床实践之前，我们需要有意识地应对幻觉并保护患者的机密性。我们提出了这些缺点的可能补救措施，并强调了人类交互在其发展中的关键作用。
这些工具的整合有可能彻底改变泌尿外科，但也提出了几个需要关注的挑战。为了充分发挥这些模型的潜力，泌尿科医生必须始终将其临床意识和经验融入到对这些工具的培训中。我们敦促泌尿外科界积极参与 AGI 和 LLM 的开发，以应对潜在的挑战。这些模型可以帮助我们充分发挥潜力，帮助我们实现更好的工作与生活平衡。
