Man, Machine, Toothpick

The Games People Play

© Photo by Bertie

I try to read one medical journal article a day. My favorites are the Case Records of the Massachusetts General Hospital in the New England Journal of Medicine, where a master clinician is presented with a baffling case, walks through the reasoning, and then learns whether he was right.

This week’s case had a twist: the human doctor went head-to-head with artificial intelligence.

Both were given the same case—no extra clues, just the clinical facts—and asked to figure out what was going on. The patient was a 36-year-old man with fever, abdominal pain, and shortness of breath. Scans showed cloudy lungs, swelling around the liver and intestines, and a clot in the vein leading from his right kidney. Blood tests revealed infection with Streptococcus anginosus, a bacterium that forms deep abscesses.

The human diagnostician, Dr. Gurpreet Dhaliwal, pieced it together like Sherlock Holmes. He concluded that the man’s drinking might have dulled his reflexes and led him to swallow something sharp. When doctors finally looked with a scope, they found a five-centimeter wooden toothpick that had pierced through the intestine, causing the infection that spread to the lungs and bloodstream.

The AI, nicknamed Dr. CaBot, came remarkably close: it diagnosed a perforated ulcer in the intestine that had led to an abscess, a blood-clot infection, and sepsis. It missed the toothpick—but not by much. And if you ask me, the machine’s reasoning made more sense: what’s more likely in an alcoholic, an ulcer or a mistakenly swallowed toothpick?

From this strange and fascinating contest, I had three disturbing thoughts.

First: the real story isn’t who won. It’s that the machine already performs better than all but the most elite humans. It reasons faster, recalls more, never tires, and can be carried in your pocket. The question of whether AI will surpass the greatest clinician is a moot point— I’m sure it will; but it’s already surpassed what matters more: the rest of us.

Second: as AI gets better, most of us will find ourselves replaced or redeployed. Already in intensive-care medicine, one specialist can remotely monitor multiple ICUs through a wall of screens and sensors. What once required a team of experts at every bedside can now be “supervised” from afar. The lucky few will get to supervise the machines; the rest will watch from the sidelines.

Third: the same hierarchy will apply to patients. Most people will be treated by machines supervised by humans—avatars on a screen offering algorithmic empathy. A lucky few will be treated by humans aided by machines—a real doctor with a face, a heartbeat, and a touch, supported by technology that sharpens judgment and reduces error. That version of care will become a luxury item.

I am reminded of Garry Kasparov’s famous chess match against IBM’s Deep Blue. I was actually there—in Philadelphia, with my friend Peter from medical school—watching history unfold. He claims I mocked it at the time, but we both knew we were witnessing something momentous.

And here we are again, almost thirty years later, seeing the same story play out - only orders of magnitude greater - in medicine, in law, in finance, in the arts, in every field where we thought we had a unique human advantage.

Kind of a buzzkill, but here’s a thought.

As machines master one game after another, the challenge is no longer to beat them—it’s to rediscover the ones that only we can play. Games where a human soul is the ticket to admission. We may even end up thanking the machines—not for beating us at the games we’ve been playing until now, but for pointing us toward the ones we should be playing all along.