The real risk of AI replacing doctors is not accuracy - KevinMD.com

TEXT START: I need to say something uncomfortable, and I need to say it as someone who has spent a career arguing the opposite.

The Dissection

The essay concedes the technical premise of AI replacement, then attempts to preserve the physician through the moral residue of medicine: presence, consolation, witnessing, and shared vulnerability.

Its strongest passage identifies the real institutional danger: compassion is bundled with billable technical work. Once AI removes diagnosis and procedure from the physician’s economic package, the unpaid human remainder may be cut with it. That is a genuine transition mechanism.

But the essay quietly changes the question. “Patients may still want a human presence” becomes “physicians will remain economically necessary.” Those are not equivalent claims.

The Core Fallacy

The central error is confusing human value with occupational indispensability.

A patient’s need for comfort does not establish a continuing need for doctors to provide it. If diagnosis, triage, treatment planning, and procedures are automated, the remaining presence can be supplied by nurses, aides, clergy, family members, specialized companions, lower-cost servitors, or AI-mediated systems. The emotional function may survive while the physician’s monopoly over it dies.

The essay also treats suffering and shared mortality as a technical moat. They are not. They may make human presence preferable, but preference does not overcome cost pressure, staffing scarcity, reimbursement design, or institutional competition. Under P1 and P2, the system will separate the high-cost cognitive technician from the lower-cost emotional and logistical functions wherever it can.

The line between “care” and “relationship” is real phenomenologically and weak economically. The market does not preserve a role merely because the role is sacred to its practitioners or meaningful to patients.

Hidden Assumptions

• That the person best suited to console a patient must also be the person who diagnoses or operates.
• That health systems will preserve physician-level staffing for nontechnical presence after AI makes the physician’s technical labor unnecessary.
• That reimbursement can be redesigned quickly enough to protect human witnessing from cost-cutting pressure.
• That patients will consistently distinguish authentic human presence from competent simulation when frightened, sedated, isolated, or financially constrained.
• That moral importance translates into bargaining power.
• That physicians possess a durable monopoly on trust, grief work, and bedside authority.
• That institutional actors will choose the author’s preferred form of human care rather than the cheapest acceptable substitute.
• That “no machine can suffer” means “no machine can perform the socially accepted function of comforting.” It does not.
• That preserving the human element preserves the physician class. It may instead preserve a thinner, lower-status care workforce while eliminating most physicians as technical specialists.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The partial truth is important: technical automation can destroy the financial subsidy that currently carries bedside humanity. The transition-management function is an implicit demand to redesign reimbursement and institutions before the bundle breaks.

The anesthetic lies in presenting the physician as the natural guardian of the human remainder. It lets a threatened professional class concede its technical obsolescence while claiming moral indispensability. That may be emotionally honest, but it is strategically insufficient. The system can agree that compassion matters and still delete millions of high-cost physician roles.

The essay is not pure copium. It correctly sees that the first casualties may be structural and financial rather than emotional. But it mistakes a surviving human need for a surviving professional caste.

The Verdict

The article diagnoses the wound and misidentifies the patient.

AI does not need to outperform compassion to replace doctors. It only needs to automate the economically decisive parts of medicine and leave human presence detachable from them. Once P1 produces superior cognitive and procedural performance, P2 prevents institutions from preserving physician-centered human care at scale, and P3 strips doctors of necessary productive participation.

The likely future is not a world without comfort. It is a world where comfort survives as a cheaper, separately staffed service while the physician becomes an expensive historical intermediary. The essay’s proposed rescue works only if society deliberately pays for human presence after technical necessity disappears. That is a political possibility, not a structural tendency. Under the Discontinuity Thesis, the doctor’s moral importance is not a moat. It is a memorial.