The Radiology Lesson
I drive to Statesboro for work, it is an hour-long drive, enough to listen to podcasts but also reflect on the changes in education, specifically as it applies to higher education but also K-12 schools (especially as a parent: “what will my son do in the future?, a question that comes to mind a lot.
Today, I was listening to Guy Raz’s How I Built This podcast, his interview with Jensen Huang. The topic at hand was close to a topic we have been writing about recently: the impact of AI on teachers’ profession. Of course, Huang didn’t talk about teachers, but the conversation came to jobs and AI. He brought up a good example about radiologists.
He recalled, well-known experts declared that radiology would be the first profession wiped out by AI. The reasoning was straightforward: the earliest AI models were built for computer vision, computer vision could be trained to read radiology scans better than any human, and therefore no radiologists would be necessary. Huang’s response was striking: he said they were absolutely right. AI has completely revolutionized radiology. Every radiologist now uses AI to study scans, and it does it faster and more accurately than a human can.
But here is what they missed. A job has a purpose and a task. This distinction was something I knew about, of course, but hadn’t thought of it that way before. Reading scans is a lower-order task — it is routine, pattern-based, and exactly the kind of work AI excels at offloading. But diagnosing disease is the higher-order purpose: it requires clinical judgment, context, and the ability to integrate information that no model sees. AI took over the lower-order task, and in doing so made the higher-order work more visible, more valued, and more in demand. Because scanning became cheaper and faster, hospitals started ordering more scans to do a better job of diagnosing. They saw more patients, generated more revenue, and needed to hire more radiologists, not fewer. The demand went up.
Then Huang added something I haven’t been able to stop thinking about. The real harm, he argued, was not the technology but it was the prediction itself. When well-known experts declared the end of an entire profession, young people who might have gone into radiology decided it was obsolete and chose a different path. The result: the world does not have enough radiologists. The narrative did what the technology could not.
I believe the same dynamic is unfolding in teaching.
The dominant story right now frames GenAI as a replacement for instructional design, assessment creation, and feedback — the lower-order production tasks of teaching. And it is true that AI can handle much of this production. But production is not the purpose of teaching. The purpose is learning: designing experiences that spark curiosity, building relationships that sustain effort, mentoring the kind of thinking that no prompt can scaffold. That is the higher-order work. And just as in radiology, offloading the lower-order tasks should make the higher-order work more central, not less.
If we tell the replacement story loudly enough, two things happen. Practicing teachers start ceding the higher-order work to the machine — not because the machine does it better, but because the narrative says it should. And prospective teachers never enter the profession to develop those higher-order skills in the first place.
That second risk is the one that keeps me up at night. We are not short on AI tools for education. We are short on people who know how to teach well. Every person discouraged from entering the profession by the “AI will replace teachers” headline is a person who will never learn to design a lesson that lands, read a room that is struggling, or ask the question that changes how a student sees themselves.
AI will change teaching. It already is. But the profession will not be eliminated but it will definitely be reskilled around the work that matters most. The question is whether we tell that story clearly enough to keep talented people walking through the door.