The Doctor and the Algorithm. Promise, Peril, and the Future of Health AI
AI is a clear and present danger to health, safety, and equity. AI also has the potential to improve clinical care profoundly. Both of these statements are true, and both are false by dint of their incompleteness. This kind of indeterminacy is a common problem in medicine. Famously, pharmakon (the Ancient Greek word at the root of pharmacy) means “drug” but can connote either cure or poison. The Paracelsian maxim that “the dose makes the poison” is likewise a common, albeit misleading, trope of introductory pharmacology. In many ways, AI is a pharmakon. It can be both cure and poison. The indeterminacy of a pharmakon is inarguably a challenge for medicine, but it does not bring healthcare to a halt. Rather, doctors, researchers, and regulators have slowly built up, over the centuries, systems of checks and balances that ideally lead toward more pharmakon-qua-cure than pharmakon-qua-poison. Please do not misunderstand me. This has certainly not been some sort of steady progression toward a better world. I am not trying to sell a story about the inevitability of scientific progress.
