Artificial intelligence (AI) is often described as the fourth industrial revolution, with the capacity to reshape society in much the same way as the steam engine in the 18th century, electrical power in the 19th, and computers and the internet in the 20th century.
What may set AI apart is its fundamental challenge to what it means to be human. Is it intelligence? Is it consciousness? Is it creativity? Kindness? Empathy? Courage? Is it something inexpressible?
The practice of medicine combines scientific and artistic elements. Physicians apply their technological and didactic skills as well as qualities like judgment, curiosity, empathy, compassion, and humor to help people heal. While the scientific aspect is crucial, patients often view the art of medicine—which is harder to define but just as important—as equally essential.
When asked if AI will take over physicians’ jobs, most people say no—albeit a qualified no. “AI is not positioned to replace those aspects of medical practice that are deeply human and relational,” said Emre Sezgin, PhD, principal investigator and head of Intelligent Futures Research Lab at the Center for Biobehavioral Health and assistant professor of pediatrics at The Ohio State University College of Medicine in Columbus. “The nuanced understanding of a patient’s history, the empathy conveyed in a difficult diagnosis conversation, and the ethical dilemmas often faced in treatment decisions are aspects that AI models are incapable of addressing.”
The qualified part of that answer is that although AI may not replace physicians’ central role in healthcare delivery, it will increasingly take on tasks to augment or complement what physicians currently do. Dr. Sezgin sees certain tasks as well suited for AI integration, such as analysis of large datasets, pattern recognition, and routine administrative tasks. “AI-powered tools can enhance diagnostic accuracy via such support, optimize personal treatment plans, and handle administrative duties,” he added.
In a commentary published in Digital Health, Dr. Sezgin described AI as repurposing, not replacing, physician roles. He sees a collaborative relationship between AI and physicians, often referred to as a ‘human-in-the-loop” (HITL) approach, in which human judgment is integrated into the loop of AI’s decision-making or learning process; eventually AI interacts with humans to augment care. With human–AI collaboration, for example, AI can offer insights into clinical decision making that physicians can then leverage with their own knowledge to make clinical judgments. The feedback AI gets from the ways physicians make their final decisions based on AI input helps to improve AI adaptability. Such continuous engagement between AI and humans can facilitate greater diagnostic accuracy and clinical decisions.
Dr. Sezgin underscored the idea that this type of relationship between AI and clinicians “requires a shift in mindset for continuous learning and adaptability.” For clinicians, this means staying informed about advancements in AI and understanding how to integrate the technology into clinical practice to enhance patient care.
Otolaryngologists Are Insulated from AI Takeover, but Change Is Coming
Most otolaryngologists interviewed for this series agree that AI is unlikely to take jobs from otolaryngologists, particularly given that otolaryngology is a surgical specialty. Image-intensive specialties, such as radiology, pathology, and dermatology, are expected to face greater AI takeover (Mohta A. These 7 specialties may be obsolete in the next decade. MDLinx. October 23, 2023). But, as described in this article series, AI is very likely to augment tasks that currently bog clinicians down in time-consuming requirements that detract from patient care and can lead to burnout.
Alfred-Marc Iloreta, MD, assistant professor in AI and emerging technologies in the Graduate School of Biomedical Sciences at Icahn School of Medicine at Mount Sinai Hospital in New York, said that there’s potential for AI to take on repetitive tasks such as scheduling patients, dictating notes, and compiling billing. “I think the best way to view AI is that it represents an incredibly powerful tool that places a broad amount of expertise at our fingertips in a very immediate and convenient fashion,” he added.
But, like any tool, he said, AI is limited by the person using it. “It will not transform a novice surgeon into a master surgeon but, more likely, will improve the learning curve and, at the same time, make that transformation safer, [more] well-rounded, and alleviate some of the collateral damage that comes with the journey, such as burnout and sacrificing extensive amounts of time,” he added.
Anais Rameau, MD, MSc, MPhil, assistant professor and the director of new technologies in the department of otolaryngology–head and neck surgery at Weill Cornell Medical College in New York, also thinks otolaryngology will face a lower threat from AI than some other specialties might, because it is a surgical specialty and she doesn’t see the technology taking over the procedural skill set. “These tools are definitely not here to replace us right now,” she said. “We want to have humans work alongside AI and not have AI make autonomous decisions.”
We are behind in our understanding of AI and, unfortunately, we have a lot of catching up to do as a community in understanding what machine learning is, what large language models are, how we can benefit from these technologies, incorporate them, and evaluate them critically.” — Anais Rameau, MD
In otolaryngology, Dr. Rameau, who is also associate editor of machine learning for Laryngoscope and a member of the AI Task Force of the American Academy of Otolaryngology-Head and Neck Surgeons (AAO-HNS), sees AI as augmenting some tasks, like documentation, that clinicians find burdensome, as well as being used as decision support tools to assist clinicians in refining their differential diagnoses or specific diagnoses of patients. “There might be predictions and assessments that AI is able to create that an individual physician will not be able to do as well—such as incorporating complex patient data, such as genetic or socio-demographic information—to arrive at a more precise diagnosis,” she said.
Vijay R. Ramakrishnan, MD, professor of otolaryngology–head and neck surgery and director of rhinology research at Indiana University School of Medicine in Indianapolis, said the surgical aspect of otolaryngology may account for why otolaryngologists should not be particularly worried about AI taking over their jobs.
In a survey on physician views of AI in otolaryngology and rhinology published last year in Laryngocope Investigative Otolaryngology, Dr. Ramakrishnan and his colleagues found that otolaryngologists were primarily concerned about whether they could trust AI to be accurate and unbiased (particularly when AI disagreed with their clinical judgment), what their medicolegal exposure might be if they disagreed with AI and followed their own clinical judgment, and the possibility of third-party payers creating or using AI systems to deny or limit care recommendations or create roadblocks for reimbursement (2023;8:1468-1475).
Despite these concerns, the survey showed that most otolaryngologists were very receptive to AI for clinical use and particularly excited about using it for clinical decision support, for personalized medicine, or potentially for the diagnostic or therapeutic management of patients. “The otolaryngologists we interviewed were all very hopeful that these uses will enhance patient care rather than replacing our roles,” said Dr. Ramakrishnan.
The most striking finding of the study, he said, was that most of the otolaryngologists surveyed had very limited understanding or hands-on experience with AI applications. The survey was conducted prior to the release of ChatGPT and, since then, Dr. Ramakrishnan said he thinks most people have some awareness of the basics of AI and its developments in healthcare.
Practicing Medicine in the Age of AI
For Dr. Rameau, who is board certified in clinical informatics, it’s critical that otolaryngologists become familiar with, educated about, and unafraid of AI to prepare for the inevitable integration of AI applications into their practices. “We are behind in our understanding of AI, and, unfortunately, we have a lot of catching up to do as a community in understanding what machine learning is, what large language models are, how we can benefit from these technologies, incorporate them, and evaluate them critically,” she said.
Her colleagues in radiology are saying that those who are using AI will have jobs, while those not using AI may not. “This could be true for otolaryngologists,”
she said.
Dr. Rameau encouraged otolaryngologists to work alongside engineers who are developing these tools to make sure the tools are applicable to otolaryngologic care. Noting that few otolaryngologists are board certified in clinical informatics, she underscored the need for more members to earn a certification in this specialty area that provides an in-depth understanding of issues such as interoperability and machine learning.
For all otolaryngologists, she urged fearlessness in embracing change as they experience the excitement and exercise the caution needed when adopting a powerful tool such as AI.
Dr. Ramakrishnan also encouraged otolaryngologists to actively participate in developing, testing, and directing AI applications to ensure they offer needed solutions for the specialty. “I am concerned that the growing hype [around AI] is resulting in AI being a hammer that is looking for nails,” he said, “and that AI applications in healthcare will be most rapidly developed by large well-funded groups for the business of healthcare rather than the delivery of it.”
There are aspects of medical care that will always require a human physician.” —Maya G. Sardesai, MD
Regardless of the hype, excitement, and caution around AI in healthcare, and in otolaryngology specifically, what remains at the end of the day are those parts of being a physician that machines can’t easily replace. “There are aspects of medical care that will always require a human physician,” said Maya G. Sardesai, MD, MEd, associate professor and associate residency program director in otolaryngology–head and neck surgery at the University of Washington School of Medicine in Seattle. Among these aspects are the need for human judgment to discern, for example, atypical presentations of a disease that don’t necessarily follow statistically common patterns, the need to empathetically incorporate individual patient values and cultural factors into the planning and decision making around medical care, and the personal back and forth of human conversation upon which the doctor-patient relationship is built, often over time. “This relationship itself can contribute to healing,” she said.
Epilogue
For now, at least, AI is less poised to replace physicians in tasks requiring qualities intrinsic to good medical care (observation, curiosity, judgment, empathy, warmth, and humor, to name a few) and more positioned for tasks involving rapid assessment of large amounts of data, finding patterns in that data, and generating potential answers or solutions to given questions. Whether or not AI will one day be able to effectively mimic more “human” qualities has yet to be seen, but rapid developments in foundational models being used to map the basic structural and functional units of life (cells) suggest that, one day, AI may be able to inform us of what we are as living organisms.
Stephen Quake, PhD, a biophysicist at Stanford in Palo Alto, Calif., who helped develop universal cell embedding, a foundational model trained on a collection of cell atlas data from humans and other species, said, in a recent New York Times article, that the model “essentially rediscovered developmental biology” in its ability to, for example, teach itself how cells develop from a single fertilized egg (Zimmer C. A.I. Is Learning What It Means to Be Alive. The New York Times. March 12, 2024). “I think these models are going to help us get some fundamental understanding of the cell, which is going to provide some insight into what life really is,” he said in the article.
The question of whether AI will or can replace humans, particularly in endeavors that, at their essence, require the most human of qualities to achieve their aims—like physicians administering care to their patients—may one day be turned slightly on its head if AI is able to illuminate for humans what makes us human.
Mary Beth Nierengarten is a freelance medical writer based in Minnesota.