ChatGPT Better Than Doctors?!

[@jamesbarry1673]

I'm a retired New York City hospital administrator who was admitted to Lenox Hill and I can tell you I almost signed myself out, I was so upset with the quality of care I was receiving FYI I was admitted on October 26th of this year

[@jThorSF]

For the last 7-8 years of my mother’s life, I moved into the hospital with her during the 7-9 stays she had (strokes mostly). I absolutely know this prolonged her life‼️

[@joandalessandro5530]

Best thing that AI can do is formulate the patient interview. Docs do not do good histories on patients. Most of em don't even DO the history. It's done by a form in the waiting room and then gone over by the med assistant and glanced at by the doc. Taking a thorough history is the most important step in diagnosis.

[@stevephillips1635]

Let’s say, hypothetically, that AI will reduce the cost of providing healthcare by 75%. Great! But is the price of using that healthcare going to drop 75%? I doubt it. This is just one example of how profits will concentrate upward with the use of AI, while jobs will disappear and the vast majority of people will be left out in the cold. AI has the potential for dramatic improvements and huge cost savings, but the question is- who will benefit financially? The few, or the many? The issue of rapidly increasing income disparity needs to be addressed, and the sooner the better.

[@benjaminlarkey8562]

I am a nurse in a rural area. Trust me. I am ready for AI. Awesome show!!!

[@user-hw4nr7ue9h]

ChatGPT is definitely better than our (burned out and squeezed by professional regulation) psychotherapists

[@litojonny]

dude sounds like adam from 'adam ruins everything'

[@Jm-Gonz]

As a recently retired PCO, by the time retired more than 50% of my entire work day was as a medical transcriptionist just wasting my time typing into Epic medical records charting to musty the patient encounter and legally qualify to submit the charge to the insurance company for the encounter I would love for a patient to type in all of their own medical history, meds, allergies, surgeries, family history, social history, symptoms, And then let Chat grind out all the research, review the labs and X rays other test results and give me the top 5 diagnosis . I could then actually spend the majority of time talking to, listening to and examining patients

[@squamish4244]

It's not hard to be better than a GP. ChatGPT wouldn't have addicted me to a massive dose of narcotics - and then been an a**hole about it and blamed me for it, tried to shirk responsibility etc.

[@ColderfromBonet]

Casey is actually handsome. We diagnosed him with my friend recently😊

[@threeofeight197]

So. For the without a doctor group did the patient interact directly with GPT 4 or was there a designated human that entered the case info into GPT 4?

[@PCMenten]

It’s no surprise to me. The good news? AI should help reduce insurance costs. Medicare/Medicaid will benefit. Edit: next test - AI vs CDC Recommended Guidelines

[@tonyhill2318]

And the thing is the o1 model is waaaay better at this sort of thing than 4.0. And even o1 is AI in its infancy.

[@AmedeeBoulette]

Try Claude Sonnet

[@alecfotsch3533]

It seems to me thay AI is really useful when a bunch of training data is available. This makes medical and computer science good targets for it to help. However i see two glaring isses with AI taking jobs. First there is a lot of work that involves moving atoms. Every physical product comes from a physical plant with humans and machines working together to move and manipulate matter. AI may help with automation but it isn't going to go replaced a roll of film on a cheese packaging line. Its not going to build a house. Second, there are lots of industries where there isn't a lot of data and the data available requires physical investigation. Medicine is actually a good example here. A doctor asks questions then actually uses their hands or eyes to look at your body. AU can't do that. It can't take physical actions unless we start building some magical humanoid robots.

[@Gangsta0wns456]

Spot on. Of course an AI can diagnose based on a written clinical vignette. Diagnosing on the wards is much more difficult. You need to obtain and good history, actually perform a physical exam, and then you can do the thinking that the AI is doing here.

[@johnthefisherman2445]

It seems inevitable that computers will replace doctors.

[@threeofeight197]

I wonder if there is a bias for GPT being better at solving complex cases. How w web MD if I have a simple cold it will tell me I have Lupus. Would the doctors have been better in a situation where there were varying levels of complexity?

[@jasonvancleve8140]

WHY IS THAT GUY HOLLERING?

[@christopherd.winnan8701]

Also interesting fidgeting with the glasses.

[@isaacabraham613]

I am delighted but not surprised. I have been going to Dr ChatGPT since it appeared and before that to Dr Google. Saves me time and money and I would never go to a human doctor unless I had a bullet wound or a broken leg.

[@superdude8168]

If you have a white collar job, you will feel what factory workers felt like during de industrialization.

[@ClementinesmWTF]

Ok cool. I’m glad. Can we automate the dock workers now? Let’s get rid of the whole capitalism thing too while we’re at it.

[@superdude8168]

@ClementinesmWTF And that attitude is why democrats lost

[@ClementinesmWTF]

@ ok cool. Thanks for showing us all why you’re an idiot lol

[@TheReddaredevil223]

Not in reality, no. Chat gpt just assumes the patient accurately reports all their symptoms and has no genuine intuition. It's not even close to being better than a doctor.

[@robinhood4640]

The results showed that GPT did better, and we know that many doctors incorrectly deny the patients observations. Maybe assuming the patient is correctly evaluating their observations, is a better approach? If a patient takes a medication that doesn't work, assuming the patient doesn't take it, and lies, because if they had it would a have worked, is not helping the patient or getting nearer to knowing what's wrong. ChatGPT doesn't do that, it evaluates the symptoms and adds the fact that the drug given is not having the desired effect. If you reject contradictory information, you shouldn't be surprised you don't find the answer as often as someone who doesn't reject contradictory information, be it another human or AI.

[@TheReddaredevil223]

​@@robinhood4640 It's not about distrusting the patients. It's about using what they're telling you as a springboard to do some tests/evaluations. Most doctors will take a patient at their word when they say they've taken a medication unless they have strong evidence to the contrary. I would say that doctors are actually more humble than chatGPT given that chatGPT is programmed to never say "I don't know the answer" and it will intentionally make up lies, which the creators have euphemistically called "hallucinations". ChatGPT doctors will go about as well in the real world as self driving cars do when they're moved off of the most basic self-contained race tracks. It's also on the bias of the evaluators to compare human errors vs AI errors with the degree of weighting that they have. When a self driving vehicle goes wrong, it goes *very* wrong and can mow down civilians at 100 miles per hour. I shudder to think what chatGPT doctors will do to patients.

[@imacg5]

It's not even about "more human medical encounter". Do Americans have enough doctors? Are all Americans have all their medical needs met with the current resource? Can all Americans afford all their medical needs? AI should be implemented to expand the medical resources, and make them affordable to everyone. Like $1 general medical consultation for everyone. Many people check WebMD when they feel something's wrong, not because of hypochondria but because they can't afford to visit a real doctor. Start with that.

[@biobossx99]

The headline is very click baity, the original study was looking at LLM as an aid for physicians for clinical reasoning compared to conventional resources. (MD here)

[@claireconover]

yea but the result showed the AI was more accurate, literally without the doctor.

[@stephenphillips6245]

LoL...who funded the study?... Meta, Microsoft and Google ?

[@imacg5]

No. All these cases are designed examples, not real patients. It means if we know the diagnosis and make up answers for possible questions, AI performs better than doctors. We don't know what happens when real patients came in and answer the "open-ended" questions. Expert system was among the first use cases for AI when computer science came up with the concept of AI, so this kind of scenario are bound to happen. The question is how to extend the current healthcare system to serve more people with the help of AI, not how to replace doctors to serve the same number of people cheaper.

[@Brendonherring522]

MD here too. Super baity. Feed it nice neat data that no computer will EVER get out of a patient. Lmao. History taking is as important to medicine as any of our lab tests or imaging studies. Just patently irresponsible to put this out here without that stipulation upfront.

[@duo315]

@@Brendonherring522 what do you expect from a tech writer?

[@charlesly]

Diagnosis is only the first step. How will AI help people afford prescription medication and to take that medication correctly whether po, inhaled or IV?

[@alecfotsch3533]

How?

[@user-hw4nr7ue9h]

you don't need a doctor for carry out those tasks

[@alecfotsch3533]

@@user-hw4nr7ue9h My skepticism is that doctors don't just receive a full patient report and then make a diagnosis from it. Instead. Typically a patient shows up has a conversation with a doctor. Then the doctor either does tests or a physical exam to gain more information. After that, they can write all the information up that was in this report and then they consider a diagnosis. For that reason, it seems like a doctor is still very much necessary. Maybe it can make the doctor more efficient and help with more accurate diagnosis, but the collection of the actual data involves something physical, whether it's some humanoid AI robot or a human doctor.

[@fergalhennessy775]

day 1 of reminding democrats that trans rights are human rights!🏳‍⚧🏳‍⚧🏳‍⚧