ChatGPT Sent Me To The ER
Three days ago, I was one hour away from hosting eighteen people for dinner when ChatGPT told me to drop everything and go to the emergency room. It was right to do so. If I'd followed my instincts instead, it could have been catastrophic.
Two days before, I’d woken up with a sore jaw and headache. I often grind my teeth at night, so this wasn’t particularly notable. It hurt a bit more than usual, but didn’t trigger any worry.
Still, I checked with my health consultant, Dr. GPT Thinking 5. Its assessment matched mine: likely TMJ from grinding. Could be something else, but those alternatives were rare enough to dismiss. I took some ibuprofen and went on with my day.
The next day, my neck hurt worse than my jaw. Maybe I was getting sick? Sometimes my lymph nodes get sore when I'm fighting something off.
That evening, I noticed my left eye felt odd. I went to a mirror and was surprised to see my left eyelid drooping, covering much more of my eye than usual. Again, kind of weird but not totally unlike things I've had before. Maybe what I was fighting off was an eye infection?
I reported this new symptom to ChatGPT, which offered a menu of possibilities: sinus infection, conjunctivitis, cluster headache, or (less likely) something neurological like Horner's syndrome. It recommended I get checked out "soon, ideally today."
With dinner guests arriving in an hour and a four-day trip starting the next morning, my initial plan was to just power through. I'd grit my teeth through dinner, get some sleep, and hope for improvement.
You’re likely familiar with ChatGPT’s annoying tic where it ends lots of answers with "Would you like me to..." to keep the conversation going. I usually find this rather annoying, and have debated adding some custom instructions to suppress it.
However, this time it asked "Would you like me to lay out what signs would mean you should go to the ER right away?" On a whim, I said yes.
Most likely, ChatGPT explained, my symptoms pointed to something benign but uncomfortable. However, there was an uncommon but serious possibility worth ruling out: Horner's syndrome caused by carotid artery dissection. It listed six red flags to watch for.
The first two didn't apply and I ruled them out immediately.
The third item was "Unequal pupils (left clearly smaller, especially in dim light)."
"Surely not," I thought, and glanced in the mirror. And stopped. My left pupil was noticeably smaller than my right. When I dimmed the lights, the difference became even more stark.
I told ChatGPT about the pupils, and its response was "Go to the emergency department right now. Tell them you want to rule out Horner's syndrome due to possible carotid dissection."
I scrambled to make a backup plan for my inbound dinner guests, grabbed a friend for support, and headed to the ER.
When I finally say a physician, I watched his affect change as I shared my symptoms:
“My neck and jaw have been hurting for a few days.” Nothing.
“Two hours ago I noticed my left eyelid drooping a bit.” A perking up of interest.
“Also my left pupil is a lot smaller than my right.” Immediate action.
I got priority access to the CT machine, and a scan confirmed ChatGPT's suspicion: my left carotid artery had indeed dissected. The inner wall had torn, allowing blood to leak between the layers of the vessel. This created a false channel that compressed the true one, blocking most of that artery's blood flow to my brain. This is fairly bad on its own, but comes with an even greater risk of a clot forming that could travel to the brain and cause a stroke.
I was started on aggressive blood thinners and immediately transferred by ambulance to a comprehensive stroke center (having six of these to choose from is one of the perks of living in Boston).
If you've been in emergency departments a few times, you start to notice that you can tell how much danger you're in by how much attention the clinicians pay to you.
When the EMTs rolled me into the new hospital, I heard "that's him", and was descended on by the full force and power of the American medical system.
Within fifteen seconds there were, no exaggeration, eleven people in my hospital room. The speed with which I had a second IV placed, got a neurological exam, was quizzed about my symptoms, and had a plan of care was breathtaking. And moving. The urgency and focus of that many professionals directed at me made me feel relaxed and cared for in a way I'm not sure I've ever experienced.
I ended up spending two nights in the hospital, where I was given neurological checks every two hours. I was placed on new anticoagulants, and given scans with equipment that looked borrowed from a spaceship. All to ensure I hadn't suffered a minor stroke and to minimize the risk of a major one.
The verdict: my dissection was minor enough to heal on its own. My stroke risk had dropped sufficiently for discharge. Two days later, my eyelid and pupils had returned to normal, and in the intervening months I’ve made a complete recovery.
In the quiet moments that followed my discharge, I'd been thinking of something unsettling: just two years ago, before ChatGPT existed, I wouldn't have known to check my pupils. I tried googling "drooping eyelid on one side," as I would have done in the pre-ChatGPT era, and the results were so general as to be useless. None of the top hits mentioned checking pupil size as a diagnostic step.
If not for AI's guidance, it’s possible I would have suffered a stroke while sleeping alone in my apartment.
When I scrolled to the top of the chat where I’d first reported my neck pain, I saw that ChatGPT had mentioned a possible carotid dissection in its very first response to me. It started as a low probability guess that increased in certainty as my symptoms progressed. I find it striking that it was considering that possibility from my very first message.
I won't pretend I'm not concerned about some of AI's potential impact on us in the future, but here in the present it may have saved me from life-changing catastrophe.
So glad you’re okay!! This does, in my opinion, say far more about the state of general/mainstream medical knowledge (and possibly the state of our insurance system) than it does about LLMs. If you ever notice a neurological symptom: go get checked out. Period. A neuro symptom is not like a sore muscle or a tummy ache. It’s almost always a sign of something serious. I would’ve immediately gone to an urgent care or ER when I noticed the droopy eye.
I'm a harsh critic of advanced AI in general and LLMs in particular. I think there are nearly a dozen distinct kinds of existential threats that they pose to society. Given that, this anecdote is good for keeping my zealotry in check. I'm glad that you're doing well!