Listen, I’ve spent enough time in emergency rooms - both as a patient and killing time between bars - to know that doctors aren’t exactly the infallible gods they pretend to be. But here’s something that’ll make you spill your drink: ChatGPT just spanked a bunch of MDs at their own game, and I’m not talking about golf at the country club.
Let me set this straight while I pour another bourbon: Some docs at Beth Israel Deaconess (fancy name for a hospital, right?) decided to pit ChatGPT against real flesh-and-blood physicians. One guy, Dr. Rodman, thought he knew exactly how it would play out - AI would be the trusty sidekick, like my liver to my drinking habit. Boy, was he wrong.
Here’s the score: ChatGPT nailed 90% of the diagnoses. The docs? A measly 74%. Hell, even when they got to use ChatGPT as a helper, they only bumped up to 76%. That’s like having the answers to the test and still failing.
The funny part? These weren’t some weird space diseases or stuff you’d see on House. These were regular cases that have been collecting dust since the ’90s - you know, back when we thought Zima was a good idea. The AI hadn’t even seen these cases before, which makes it even more impressive. It’s like watching a rookie beat Vegas odds their first time at the table.
But here’s where it gets really interesting: The doctors wouldn’t listen to the damn machine even when it was right. Reminds me of my ex-wife’s stubborn cat that would rather starve than eat the fancy food I bought it. These MDs were so convinced they were right, they basically told ChatGPT to go screw itself.
And get this - most of them didn’t even know how to use the AI properly. Instead of giving it the whole case, they’re pecking at it like my uncle Tommy trying to send his first text message. “Hey ChatGPT, what causes a headache?” Real sophisticated, docs.
They’ve been trying to make computers diagnose stuff since the ’70s - back when bell-bottoms were cool the first time. Some guy named Dr. Myers even tried to download his brain into a computer program called INTERNIST-1. Spoiler alert: it didn’t catch on. Turns out doctors don’t like machines telling them they’re wrong. Who knew?
The real kick in the teeth? These modern AI systems aren’t even trying to think like doctors. They’re just really good at predicting what words should come next. It’s like that drunk guy at the bar who can finish everyone’s sentences, except this one’s actually right most of the time.
Now, I’m not saying we should replace all doctors with computers. Hell, I still need someone with opposable thumbs to write my prescriptions. But maybe - and I’m just spitballing here while I reach for the bottle - we should start listening to these silicon-based second opinions.
Because here’s the truth, served straight up: We’re all human, even those folks with MD after their names. And humans make mistakes. Sometimes they make mistakes because they’re tired, sometimes because they’re stubborn, and sometimes because they just can’t admit they might be wrong.
At least the AI doesn’t have a God complex or student loans to justify its existence. It just does its job, like a really efficient bartender who never judges you for ordering another round.
The next time you’re sitting in some doctor’s office, getting diagnosed with “probably nothing serious,” maybe ask if they wouldn’t mind running it by their digital colleague. Could save you a lot of trouble. Or don’t - what do I know? I’m just a tech blogger who’s on a first-name basis with his local liquor store owner.
Time to ice this headache. ChatGPT says it’s dehydration, but I’m pretty sure it’s Tuesday.
Yours truly from the bottom of the bottle, Henry Chinaski
P.S. If any doctors are reading this, don’t get your scrubs in a twist. We still need you - someone’s got to click the mouse, right?