Artificial intelligence (AI) has permeated virtually every aspect of our lives. AI has become the acronym of 2023, and folks just love acronyms; in particular, those that are culturally and socially magnetic. From LOL to F2F to TGIF, these shortened representations are used repeatedly and frankly, often incorrectly. The term AI is used by many to start a conversation without an understanding of what it really is, how it is generated, and what it can and what it cannot do. From those who casually dismiss it as a 2023 phenomenon to others who believe it is the beginning of the apocalypse, it behooves us as scientists and patient caretakers to learn more about AI and embrace its advantages and disadvantages. Essentially, AI is a superfast metadata analyzer that, as the authors explain, can “categorize, classify, learn, and filter data to solve problems, make predictions, and perform other seemingly intelligent tasks.”¹ In particular, the ability to harness enormous amounts of data and differentiating trends can help expedite research and innovation direction. There are also subcategories including machine learning, where machines are given data access to learn on their own, and deep learning, which employs neural networks. But like other types of data analyses, even those by our lower brain processes, the quality is based on the data incorporated into the model. The classic “dirty data in, dirty data out” has been the historic challenge in interpreting large volumes of information. Bias, intentional or not, with respect to what AI digests can manifest in multiple ways. Exclusion of datasets that are more broadly applicable and inclusion of data that reflect only certain disease states or populations render AI conclusions less meaningful.

But let’s not underestimate the power of AI. Bing AI and ChatGPT4 have an IQ of 114 and are smarter than the average human. By comparison, the average human IQ is 98, and 68% of humans have an IQ of between 85 to 115. Yet, challenges remain. For example, there are various applications of racial and ethnic correction factors in clinical data and these should be incorporated in a more meaningful way.¹ Sometimes we may overcorrect as opposed to looking at these datasets as artifacts and letting AI identify what is missing rather than just figure out solutions by only what exists. In essence, data inconsistencies cannot be simply overcorrected by a technical or software solution. This is true for racial perspectives, missing data, and population disparities. In addition, McKinsey reported that the most cited risks with AI are inaccuracy, cybersecurity, and intellectual-property infringement.²

That being said, and with the preceding caveats noted, urology is ripe with focal points for AI analysis. For example, in benign prostatic hyperplasia (BPH), areas to focus on include: (1) imaging where AI can be used to analyze medical images of the prostate, such as MRI or ultrasound scans, to help diagnose BPH and assess its severity, (2) predictive analytics where, with data from electronic health records, AI can help predict which patients are at higher risk of developing BPH, allowing for early intervention and treatment, (3) treatment planning where AI can help physicians plan and optimize treatment for BPH, taking into account factors such as patient age, severity of symptoms, and other health conditions, and (4) using AI to assist surgeons during minimally invasive procedures for BPH, such as transurethral resection of the prostate, helping to improve the accuracy and safety of the procedure.

This is only a sampling of how AI can impact urologic discovery and research. All of the disease states within urology can, and will, be impacted by the emergence of AI, and no doubt we will begin to see the paper polymerase of articles and studies begin to appear. My best advice is to get educated about AI and avoid FOMO!