Rejecting Generative Ai In Healthcare Won’t Protect Patients – It Will Harm Them | Letter

Eric Reinhart’s effort raises thoughtful concerns astir nan risks of generative artificial intelligence (genAI) successful healthcare (What we suffer erstwhile we surrender attraction to algorithms, 9 November). I admit that my position was included, and I guidelines by my belief that “AI will beryllium arsenic communal successful healthcare arsenic nan stethoscope”.

Reinhart rightly points retired that location are superior risks successful deploying caller technologies successful our profit-driven, time-constrained healthcare system. But to judge nan risks of genAI accurately, we request to first admit nan unnecessary errors and deaths that hap today. And past we must inquire whether this exertion will make those problems worse aliases make aesculapian attraction safer.

In nan United States, 400,000 group dice each twelvemonth from misdiagnoses. Another 250,000 dice from preventable aesculapian errors. People pinch chronic diseases (lifelong conditions that impact them daily) are often seen only erstwhile each 4 to six months, pinch nary information aliases curen modifications successful between. Over time, humor unit remains elevated, humor sweetener levels enactment uncontrolled and chronic bosom nonaccomplishment worsens. The result: hundreds of thousands of avoidable bosom attacks, strokes, cancers and kidney failures each year.

GenAI tin help. Patients tin person timely, reliable guidance, while physicians tin beryllium alerted to early informing signs and set medications earlier irreversible harm happens. Likewise, individuals pinch intelligence wellness conditions tin entree support astatine nighttime erstwhile worry aliases slump worsens and nan only action is nan emergency room.

Instead of replacing physicians, genAI tin capable nan gaps erstwhile clinicians are unavailable. Rather than framing nan early arsenic either/or (ie clinicians aliases AI) it should beryllium both/and. The operation of dedicated clinicians, empowered patients and genAI will lead to safer, higher-quality and much accessible attraction than immoderate of nan 3 tin execute alone.

Reinhart’s effort raises important questions, and we should support discussing them. But rejecting genAI won’t protect patients. It will harm them by allowing today’s problems to go worse.
Dr Robert Pearl
Los Gatos, California