Artificial Intelligence

Compassion. That’s the one thing no machine ever had. Maybe it’s the one thing that keeps men ahead of them. — Leonard McCoy, MD (2268)
In March, the Centers for Medicare & Medicaid Services (CMS) announced the start of its Artificial Intelligence Health Outcomes Challenge. This program is intended to accelerate the development of artificial intelligence solutions (AI), and “help front-line clinicians understand and trust artificial intelligence-driven data feedback to target scarce resources and improve the quality of care.” It may sound like science fiction, but the prospect of artificial intelligence assuming a greater role in healthcare is very real, and its advent is surprisingly not too far away.

Today, artificial intelligence can drive autonomous vehicles, and beat the smartest humans on Jeopardy! Many experts predict that artificial intelligence (AI) will progress to the point where machines will be able to perform just about any job better than a human. It’s already happening. Siri can route you to an unfamiliar location better than any human with a paper map. Google Translate has reached the point where it rivals human interpreters. And Amazon is alarmingly good at showing me advertisements for things I didn’t know I desperately needed.

While computing power continues to grow exponentially, future developments will allow computers to advance beyond the brute-force execution of predefined algorithms. With machine learning, AI will be able to interpret problems and implement solutions in much the same way our brains work—but faster and with greater reliability.

AI will be able to take on many complex tasks, but can a robot practice medicine? The short answer is yes. It won’t happen all at once, but it is inevitable. Some experts, like Sun Microsystems co-founder Vinod Khosla, predict that robots will be able to replace 80% of doctors by 2035. Students entering medical school this year will have barely begun their professional careers when this brave new world arrives. My advice to a student embarking on a career in a procedure-based medical specialty is to consider the possibility that they may be able to practice only 6-8 years before a robot arrives to take their place. This will leave precious little time to earn enough to pay off their student loans.

We have seen this story before. It happened at the dawn of the industrial age, as craftsmen who worked with their hands were gradually replaced by more efficient machines. Just as it was with the 19th Century textile industry, there will be many who resist the rise of automation. And just like their Luddite ancestors, they will fail to stop progress.

At first, the transition will be subtle. Speech recognition will eliminate medical transcriptionists. Computers will be put to work interpreting X-rays and pathology reports. Big data companies will condition the public to accept medical advice from their online services. The government will announce grants to accelerate the development of—wait, all of that has already happened!

It’s already happening, and the AI revolution will only accelerate. AI will likely replace many of our procedure-oriented colleagues during the next few decades. It’s easy to imagine a future in which we refer our patients with gallstones directly to a robot surgeon rather than to a human surgeon who does robotic surgery. Surgical skills and test interpretation are very conducive to automation, but what about the cognitive skills of family medicine? Surely what we do is too subjective and complex for automation. We shall see.

Our AAFP President, John Cullen, spoke on this topic at the 2018 Family Medicine Experience. In his view, AI will eventually replace all physicians. One day in the distant future, as Dr. Cullen imagines it, the last doctor will hang up her stethoscope. If there is any consolation, he predicts the last doctor will be a family physician.

I like to imagine that family physicians will be among the last 20% of the workforce that can’t be replaced by a machine. Maybe, instead of stealing our jobs, AI will make our jobs easier. Imagine a future without typing or box-checking. That’s the optimistic view of the AI revolution. It would be nice to have a self-documenting medical record, an AI that listens in on the patient visit, types up the note, and offers friendly advice on disease management. It would also be nice to have a robot to handle all the messages in my inbox and tell me only what I needed to know. This is the future I would like to see.

I suspect that when future family physicians hang up their stethoscopes it will not be because family medicine is obsolete. Instead, we will hang up our stethoscopes because we have replaced them with more effective tools. Like Dr. McCoy from the old Star Trek series, the mechanical trappings of medical care will have been replaced with diagnostic beds and handheld scanners. It’s possible. After all, Star Trek was remarkably prescient in depicting a future filled with voice-activated computers, digital display pads, and ubiquitous communicators. Shouldn’t we also trust its depiction of 23rd Century Healthcare?

The above introductory quotation comes from a 1968 episode of Star Trek, “The Ultimate Computer,” one of many episodes that deal with the perils of automation and artificial intelligence. (If you haven’t seen it, just say “Alexa, play Star Trek, season 2, episode 24). Predictably, Dr. McCoy is the voice of reason, warning Kirk and Spock that an over-reliance on computer technology will lead to disaster. He is proven correct, and in the end, it is human compassion that saves the day. In many ways, Dr. McCoy should be our role model for the future family physician—a generalist, willing to use technology, but ever skeptical of losing the human touch.

Grappling with the role of technology in health care also reminds me of a very real hero of medicine, Archie Cochrane, one of the fathers of evidence-based medicine. In his autobiography, One Man’s Medicine, he writes of his experiences as a prisoner of war. In 1943, he was the sole medical officer working in a tuberculosis hospital in Elsterhorst, Germany. One night, the Germans delivered a Soviet prisoner, “moribund and screaming” in pain from cavitary tuberculosis. Unfortunately, Cochrane had no technology with which to treat his patient:
“I felt desperate. I knew very little Russian then and there was no one in the ward who did. I finally instinctively sat down on the bed and took him in my arms, and the screaming stopped almost at once. He died peacefully in my arms a few hours later. It was not the pleurisy that caused the screaming but loneliness.”
Artificial intelligence holds the promise of delivering greater efficiency, accuracy, and safety to medical practice—and these are worthy goals. An AI physician would have had no difficulty translating Russian to English and would have been able to provide Dr. Cochrane’s unfortunate patient with a prompt and accurate diagnosis. However, sometimes what patients really need is to know that they are not alone in this messy business we call life. That is the essence of human compassion, and I suspect it’s the one thing no machine will ever be able to provide.

This article was published in South Carolina Family Physician.