Ƶ

Beyond Memorization: AI Can Revolutionize Medical Education

— Tools like ChatGPT could catalyze the trend toward a "flipped classroom"

Ƶ MedicalToday
A photo of the ChatGPT application running on a smartphone over the OpenAI logo.

Since the -- an artificial intelligence (AI)-driven text generator developed by OpenAI -- in November 2022, it has garnered status as the web application in history. On a daily basis, users eagerly share newfound applications of the tool, from to to . Healthcare is no exception: physicians have experimented with tasking ChatGPT to summarize medical records, generate treatment plans, and even write letters seeking prior authorizations. On top of that, a has reported the ability of ChatGPT to complete the U.S. Medical Licensing Examination (USMLE) with an accuracy near or at the passing threshold.

While ChatGPT and similar tools will not be replacing clinicians anytime soon, the technology does highlight the triviality of the memorization of medical facts. What does it mean for medical education if such an AI tool can, without any special training and in an infinitesimal fraction of the time, perform at a level that takes medical students 4 years and thousands of hours of dedicated study to attain?

The performance of ChatGPT on the USMLE is a wake up call that the medical school curriculum and evaluations systems must change.

For years, leading medical educators like Charles Prober, MD, founding director of the Stanford Center for Health Education, have been advocating for a move away from traditional lectures and a memorization of facts. He for a "flipped classroom" approach to medical education, where students can gather facts and lectures on their own time, and then come to the classroom to interact with professors and peers to practice problem-solving and data analysis. In this model, medical students learn to solve complex patient cases with their computers and every resource available to them -- just like they will in the real world. This approach aims to de-emphasize the memorization of medical facts and focus on interacting with data and resources to develop critical thinking skills.

While this method of teaching is in medical schools across the U.S., ChatGPT can further facilitate this approach. Without AI tools, finding relevant pieces of information can be time-consuming and frustrating, often requiring a search through multiple sources. ChatGPT and similar models provide an order of magnitude reduction in the time it takes to find, synthesize, and present relevant facts.

More than anything, ChatGPT epitomizes the effortlessness of information retrieval in the age of AI and machine learning. The medical education community must take proactive steps to take advantage of these technological advancements. While medical education has traditionally moved at a slow and measured pace -- sticking to core concepts and deliberately incorporating new tools and information over time -- ChatGPT is not simply another new tool but rather represents a paradigm shift in what it means to learn medicine.

How can AI and machine learning tools like ChatGPT be incorporated into medical training? Just as database searches are a part of medical education today, AI tools can also be introduced to medical trainees. While some may consider using these tools as "cheating," it's important for students to engage with and understand the strengths and limitations of this technology. What kinds of medical conditions can we trust ChatGPT to diagnose? What nuances to treatment plans is ChatGPT unable to appreciate? Finally, regardless of how physicians feel about it, patients now have access to ChatGPT; it is therefore the responsibility of clinicians to understand the capabilities of these AI tools so they may play an active role in interpreting and managing the consequences of easily accessible medical -- and pseudomedical -- information.

The growing adoption of AI tools further underscores the vital role of future physicians as leaders. Given that AI has demonstrated an incredible ability to follow protocols and find correct answers for straightforward medical conditions, we need to spend more effort training physicians to be team leaders. Physicians will likely be responsible for larger patient panels, assisted by interdisciplinary teams of technicians and other clinicians. In this context, physicians must learn better skills around management and communication.

Similarly, AI tools will, in conjunction with telemedicine, triage patients with simpler medical conditions into urgent care clinics or home-based solutions. In-person primary care visits will involve increasingly complex patients approaching inpatient severity, and physicians will need to become comfortable in applying fundamental knowledge in physiology to solve edge cases in medicine. In a similar vein, medical schools ought to acquaint students with exploring the full range of uncertainties and nuances of real-world medical scenarios.

Crucially, students should also focus on learning how to navigate complex social issues that go far beyond the diagnosis and basic treatment guidelines. Medical training should prepare students to tease out the preferences and values of a patient to better inform patient-centered care plans that integrate the medically "right" answer with the patient's economic and psychosocial environment. This is where doctors of the future will show their worth and skill.

The performance of ChatGPT on the USMLE at the level of a graduating medical student demonstrates the significant potential of AI and machine learning to revolutionize medical education and practice. This revolution will be expedited through several means: competition from other technology such as and ; OpenAI's ongoing efforts to improve its language models through GPT-4; and potential future neural networks specialized in medical topics. Although some may bristle at the prospect that these tools undermine the work of physicians, we believe they instead offer an unprecedented opportunity to augment what physicians can do -- diagnose diseases with ever greater accuracy and efficiency -- and enable what they do best -- connect with patients at a human level and ultimately provide the best care possible.

is an adjunct professor at Stanford Medicine in the Department of Bioinformatics Research where his teaching and research focus on digital health and AI. Additionally, he is a Partner at GSR Ventures, a venture capital fund focused on digital health. is an MD/MBA candidate and Knight-Hennessy Scholar at Stanford University, where he teaches courses on digital health and clinician-patient communication strategy. He is a member of Ƶ's "The Lab."