爆料网

The White Coat and the Iron Ring

Remarks by Under-Secretary-General Amandeep S Gill
Commencement Speech, Graduation Day Class of 2026, Medicine and Surgery, Cattolica Università Cattolica del Sacro Cuore, Rome.?

July 3, 2026

Dean Sgambato, Professor Mantini, Professor Gambassi, members of the faculty, proud families, and since you are physicians now, so let me say it properly: doctors of the graduating class.

I should warn you at the outset that I intend to talk about artificial intelligence. I am aware of the risk. There are places these days where a speaker need only say those two letters to empty half the hall, and where a tech CEO can be booed for a sentence that begins, “AI will transform everything you do.” You are within your rights to walk out. I will only point out that the oath you swore a short while ago takes a rather dim view of leaving a procedure half-finished, so perhaps stay to the end!

Let us start with that oath – the modern version of the Hippocratic Oath. Every serious profession marks the step from the classroom into the consequence with a small ceremony, because something genuinely changes at that threshold. When you left the lecture theatre for the wards, this university placed a white coat on your shoulders. Across the Atlantic, newly qualified engineers in Canada are given an iron ring, worn on the working hand - a plain band, said to recall a bridge that fell, pressed against the skin as a daily reminder that every calculation now carries a human life. The white coat and the iron ring say the same thing the oath says: from here, your knowledge touches other people, and so it must be bound to ethics before it is trusted with anyone. That is the whole point of a profession, and it is why I want to argue this morning that the oath you just took is not a relic to be set against the new machines. It is the operating manual for practising medicine alongside them - not in spite of them.

The question everyone asks about these machines is whether they will replace the doctor. It is the wrong question. The more decisive question is what they ask the doctor to become.

One line in your oath does almost all the work. You promised to remember “that I do not treat a fever chart, a cancerous growth, but a sick human being.” Stay with the fever chart for a moment. When the bedside temperature chart entered the wards in the nineteenth century, it was the advanced data technology of its day: a clean, quantified abstraction of a suffering person, hung at the foot of the bed. The oath’s instruction was never to throw the chart away. It was to never mistake the chart for the patient. That single clause, written before the first email was ever sent, is now the whole of medical AI ethics. The model’s output is a fever chart. A very good one. It is still not the patient.

Let me take you through a few of the promises you just made, and show you that each one already tells you how to live in this new century.

You promised that your patients’ “problems are not disclosed to me that the world may know.” In your time, that promise becomes a discipline of data. The same record that can train a model to catch a tumour three years earlier can, handled carelessly, follow a patient into a loan refusal or an insurance premium. Confidentiality is no longer a locked drawer. It is a question you ask of every system you feed: does this use of data advance this person’s health, and the health of others like them - or only someone’s business? Responsible, ethical use of data is the modern shape of an ancient promise. Hold the line.

You also promised to remember “that warmth, sympathy, and understanding may outweigh the surgeon’s knife or the chemist’s drug.” Here is the good news the machines deliver almost by accident: as more of the cognitive routine is shared with software, the scarce thing becomes the human thing. Empathy. Presence. The hand on the shoulder when the news is bad. Now grant the hard part honestly. For two decades the computer in medicine mostly stole that time, turning doctors into typists and the consultation into a transaction conducted over a screen. The promise of the newer tools: software that listens to the visit and drafts the note so that you can look at the patient instead of the keyboard, is that the time comes back. Unsexy, vital. Demand the tools that return you to the bedside, and refuse the ones that chain you further to it.

You promised something braver than it sounds: “I will not be ashamed to say ‘I know not.’” In an age of confident machines this becomes a double humility. You must know the limits of your own knowledge, and the limits of the model’s - which is harder, because the model never says “I know not.” It answers fluently when it is right, and just as fluently when it is wrong. Your new clinical skill is to hear the silence the machine cannot speak. Treat it as a colleague to consult, never an oracle to obey. The oath already told you what to do: call in another’s skill when a patient needs it, and keep the judgment your own.

You promised to avoid “those twin traps of overtreatment and therapeutic nihilism.” The two traps return in new clothes. The first is over-reliance - handing the machine not just the task but the thinking, until the skill that made you trustworthy quietly erodes. This is no longer a worry; it has been measured. Last year in The Lancet Gastroenterology and Hepatology, Budzyń and colleagues reported the first real-world evidence of clinical deskilling. After a few months of routinely using AI to spot polyps during colonoscopy, the same experienced endoscopists - working without the AI - caught roughly a fifth fewer of the lesions they had reliably caught before. The instrument that made them better made them worse the moment it was taken away. The opposite trap is the reflexive refusal, the nihilism that waves away every algorithm and lets a patient go without a benefit the evidence supports. Between “let it rip” and “wall it off” runs a narrow but real path: use the machine, and keep doing the repetitions that keep your own hands and eyes sharp. Curiosity about the technology and a clear sight of its risks are not opposites. They are the same professionalism.

Now go back to that fever-chart line, because it has a second half you must never forget. The human being, it says, “whose illness may affect the person’s family and economic stability. My responsibility includes these related problems.” No model on earth can see this from the data. The algorithm reads the scan; it does not know that the patient cannot afford the bus fare for the follow-up, or sleeps four to a room, or is the sole earner for a family of six. The social and economic determinants of health live precisely in the gap between the chart and the person. That gap is your jurisdiction, not the machine’s. As medicine automates everything that can be measured, your value moves to everything that cannot.

And you promised to remain “a member of society, with special obligations to all,” and to “gladly share such knowledge as is mine with those who are to follow.” Lift your eyes past this room. The same AI that hands a teaching hospital in this city its tenth diagnostic aid could hand a rural clinic, in a country with one radiologist per million people, its first. Whether that happens is not a technical question. It is a question of whether we build a floor under the divide or let the tools pool where the money already sits.

Consider two women with the same early breast cancer: one in a European teaching hospital, one in a district town with no specialist for two hundred kilometres. The technology now exists for the second woman to be screened, monitored, and even enrolled in research without leaving her province. That is the promise of two genuinely new approaches, and they are worth your attention. Decentralised clinical trials, trials that travel to the patient through remote monitoring and local clinics, instead of demanding the patient travel to the trial, can finally enrol the rural, the elderly, the people every previous trial quietly excluded, so that the evidence we make is evidence for everyone. And the patient’s own voice, captured as what we call patient-reported outcomes, structured reports of how a person actually feels in the weeks between visits, turns out to save lives. When Ethan Basch and his colleagues had patients with advanced cancer report their symptoms from home through a simple web tool, with alerts sent to the care team, those patients lived a median of five months longer than patients given the usual care. Five months, from listening. For your oldest and frailest patients - the ones who decline in the quiet weeks between appointments, who under-report, who are easy to overlook - that kind of listening, amplified by technology, may be the most humane thing you do. Not a machine replacing the conversation. A machine that makes sure the conversation happens.

A word about the training that brought you here, because it is changing too. The historian Nils Gilman has argued recently that the modern university, research, teaching, credentialing, and the coming-of-age of young adults, all bundled under one roof - is being pulled apart by AI, now that a machine can pass the examinations and write the essays. He may well be right. But notice what he says survives the disaggregation. The capacities a machine cannot supply: judgment, the constitution of a goal, the cultivation of taste, the slow formation of a person in the company of a demanding teacher, these are exactly what serious education was always for. That is your medical training, described precisely. Your worth was never the facts you memorised; the machine holds all the facts. Your worth is the judgment built by repetition, by being wrong in front of a mentor and corrected, by the reps no shortcut can give you. So be lifelong learners of the tool: fluent, curious, unintimidated. And be lifelong guardians of the judgment the tool cannot have. Both. Always both.

So, will your role shrink in the age of these machines? The honest answer is a conditional one. It will shrink if you let the model do your thinking and call that progress. It will grow, beyond anything this profession has yet been, if you let the machine carry what can be counted and spend the time it gives back on everything that cannot.

The fever chart will soon be read faster than any human could ever read it. Your work is, and has always been, everything the chart leaves out.

Go and treat the patient, not the chart. Congratulations, doctors.