AI in medical education is rapidly reshaping how future doctors learn, prompting concerns that the traditional apprenticeship model might become obsolete.
AI can already pass medical exams, but doctors remain essential
Recent tests have shown that tools such as ChatGPT can succeed on United Kingdom and United States licensing exams. Associate Professor David Kok, a radiation oncologist and educator, said the ability of generative models to answer exam questions does not mean they will replace physicians. “They lack essential higher‑order skills, including ethical judgement,” he told attendees at the #Intelligence26 conference.
Kok stressed that AI’s capacity to regurgitate information is impressive but limited. “It’s just showing you what lots of other people would have produced if asked the same question, not what it thinks the answer is,” he explained. The distinction matters when clinicians must handle ambiguous cases and make decisions that rely on moral reasoning.
He warned that the current wave of AI‑assisted learning is already influencing assessments. “Virtually no assessment is now returned to educators that hasn’t been touched by AI in some way,” the professor noted, highlighting the need for new exam formats that are less vulnerable to automated assistance.
Benefits and risks of AI‑driven tutoring
Proponents point to personalized tutoring as a clear advantage. Kok acknowledged that “AI can augment learning” and help address individual gaps.
These issues highlight the need for curricula that teach not only how to use AI but also how to critically evaluate its outputs.
Related: Path Forward for Advancing Dementia Care
In the middle of these debates, it is reasonable to expect that medical schools will adopt a hybrid model. While AI can handle routine information retrieval, human instructors will likely focus on developing judgment, empathy, and the ability to recognize the limits of any tool. This balance could preserve the depth of clinical training while still leveraging technological efficiencies.
Changing assessment methods to protect integrity
One practical response is shifting away from take‑home essays and open‑book tests toward real‑time evaluations. Objective Structured Clinical Exams (OSCEs), vivas, and workplace‑based assessments are less susceptible to AI manipulation. Kok warned that redesigning these formats will demand significant resources, noting that “reconfiguring to become essentially fully proctored type assessments is a huge burden.”
He also called for systematic AI literacy across all levels of medical training. “Every single level of learning is going to need to have some AI literacy, and so we’re going to have to build it systemically through the system,” the professor said, suggesting that a baseline understanding of AI capabilities and limitations should be embedded in curricula.
Beyond technical proficiency, educators must instill an awareness of AI’s blind spots. The technology does not possess metacognition—recognizing its own knowledge limits—and cannot experience moral agency. Without these capacities, a system can appear persuasive yet remain unsafe, he argued.
Overall, the consensus among conference speakers was that AI will not render medical education redundant, but it will compel a substantial transformation. The goal is to harness AI’s speed and knowledge base while preserving the human qualities that define competent, ethical physicians.