If, like me, you feel you are just beginning to wrap your head around all the possibilities of artificial intelligence (AI) in health care, I look forward to your thoughts on our cover story on this topic. I also highly recommend an article in this year’s ASCO Educational Book, “Artificial Intelligence in Clinical Oncology: From Data to Digital Pathology and Treatment.” The authors take a thoughtful, clear-eyed approach to the topic, exploring a variety of AI applications that are being investigated across the spectrum of cancer care (diagnosis and prognosis, treatment selection, biomarker and drug development, modulated dosing, and more) as well as potential challenges (practical, regulatory, legal, and ethical). They conclude with a number of sensible recommendations for developing AI platforms for clinical oncology. 

 

Nowhere in the article’s dozen or so pages does it say that we can’t put the genie back in the bottle, but the message is clear: This technology is here, now. It’s not going away, and we must be prepared.  

 

AI has tremendous potential in so many areas of clinical research. It can be taught to accurately review pathology and radiology, potentially forgoing the need for uploading of imaging and potentially expensive central reviews. AI can also assist in the identification of target populations for clinical trials, and at some point AI may be able to replace control groups in clinical trials, streamlining research and bringing discoveries into practice more quickly. It can assist in the design of novel therapies and identification of new biomarkers, accelerating our goal of personalized cancer care for every patient.   

 

At the same time, a high degree of caution and objectivity is warranted. As the ASCO Educational Book article says, “At the heart of these challenges is clinical validation, where clear pathways that span well-designed preclinical studies through adequately controlled [randomized clinical trials] continue to be needed.” Along with validation, we will need to be relentlessly vigilant about patient privacy, given the significant load of patient data that these models require. 

 

Whether you are excited or skeptical (or both!), the questions this moment invites are endless. What problems can AI, deep learning, large language models, etc., help us solve that we haven’t even imagined yet? How do we apply AI applications equitably, to ensure that all patients benefit? How do we prepare medical students and trainees to practice effectively, humanely, and ethically in an increasingly technology-driven environment? And above all, how do we safeguard the provider-patient relationship and shared decision-making that is at the heart of everything we do? 

 

These are not questions for some distant, hypothetical future—these applications are already in the field, and we are only going to see them grow in scope and scale. We all need to stay informed and engaged on the topic of AI in order to be the best advocates for ourselves, our profession, and our patients.