If we’re comparing worst-case scenarios: ChatGPT might confidently hallucinate a very valid-sounding explanation that convinces the patient they have a rare disorder that they spend $1000s of dollars testing for, despite real doctors disagreeing and test results being negative. Maybe it only does this once in a long series of questions, but the seed is planted despite negative test results.
Then when the patient asks ChatGPT if the tests could give false negatives, it could provide some very valid-sounding answers that say repeat testing might be necessary.
This isn’t a hypothetical. It’s currently happening to an old friend of mine. They won’t let it go because ChatGPT continues to give them the same answers when they ask it the same (leading) questions. At this point he can get ChatGPT to give him any medical answer he wants to hear by rephrasing his questions until ChatGPT tells him what he wants to hear. He’s learned to ask questions like “If someone has symptoms _____ and ____ could they have <rare disease> and if so how would it be treated?” A real doctor would see what’s happening and address the leading questions. ChatGPT just takes it at face value.
The difference between ChatGPT and real doctors is that he can iterate on his answer-shopping a hundred times in one sitting, whereas a doctor is going to see what’s happening and stop the patient.
ChatGPT is an automated confirmation bias machine for hypochondriacs.
Ah, yes. Hypochondriacs. I was a hypochondriac for years until I was able to get an appointment timed such that my symptoms were physically present while I was being assessed (not easy if you have a disease that comes and goes). I really hope you’re not a medical professional that interfaces directly with patients.
I was a hypochondriac for decades. I was eventually cured after decades of cropping hypochondria through self-diagnosis of the actual medical condition I had that became formal diagnosis eventually resulting in treatment through some modest lifestyle changes.
I'm a BIG supporter of using things like ChatGPT, google, and sci-hub to do your own medical research because the whole system where some physician diagnoses you based on an extremely limited amount of data collected in a haphazard manner after a few minutes of observation because he's experienced and smart or whatever is incredibly dumb. The way people hold it up as the ethical standard which we cannot deviate from because it would be too dangerous is utterly baffling to me. The status quo totally lacks ethics and mostly serves to line the pockets of a cartel of doctors with a monopoly on access to medication and treatment who often condescendingly think patients are simply too irrational to treat themselves without their help.
I legitimately cannot wait for this field to mature and medical self-help with AI assistance becomes the norm.
Fellow hypochondriac here. I was at the point where doctors, hospital staff, and lab techs would immediately warn new practitioners about me so they wouldn't waste finite medical resources in a small town, and I just completely discontinued normal activities out of terror as a result.
When I finally blacked out and fractured my spine, first responders detected a lifelong cardiac arrhythmia in the back of an ambulance. Only with that knowledge have I been able to receive treatment and begin to heal emotionally from the gaslighting and medical abuse I experienced while in the care of licensed professionals.
AI-assisted medicine will prevent so many of these mistakes in the future. It can't come soon enough as far as I'm concerned.
Wow, the doctors that tried to block you out completely within their network of friends must have balls of steel, what with zero fear of legal repercussions!
It was more a matter of how misinformation malignantly spreads I believe.
I had to see the only available PCP to see anyone else. This automatically prompted medical releases that, even if unethical, would have still made everything technically legal. If it was an emergency room trip, there were always the same two or three physicians there, so they all became aware of me from the first couple of episodes and could warn any specialist they referred me to see.
Same deal with laboratories and radiological facilities. When you've got only one or two options in town, they have your consent to release PHI by default if you ever want the results interpreted, and their interpreting physician can just accompany the report with a courtesy call to the receiving provider about a suspected diagnosis.
I agree with studying the field to help you understand your own health, but I prefer sci hub or any peer reviewed source over an LLM. I'll revise this view as LLMs develop, but right now I'm seeing plausible bs as often as I see good advice.
I was the same way with this while I travel--it's definitely the future. I'm working on AI healthcare assistant where you can summarize a conversation with our plugin on ChatGPT or bot on WhatsApp and then send it to a real doctor to continue the conversation.
I hope that more founders build and innovate in the field to provide efficiencies throughout the whole system to lower costs and provide high quality care for everyone that needs it.
Some insurance providers need a primary care physician for referral, but some do not, so one area we're doing research is if we can do referrals through doctor follow-up/verfication from summary of chat.
So do you not believe actual hypochondriacs exist or something? They do.
Especially the people that will read about diseases and get anxious they have them, and while you can validly say they have anxiety issues they don't have whatever they just read about 99% of the time.
Even if there's a real disease of some sort, you don't want to diagnose with the latest guess in someone that keeps guessing different things. Their treatment needs improvement, but confirmation bias is not how you do it.
It has a lot to do with whether "ChatGPT is an automated confirmation bias machine for hypochondriacs" is a valid worry or something that should disqualify you from being a medial professional dealing with patients.
Early this year I had cold symptoms that just could not get away. I was getting sick, recovering and getting sick again. After some 2 - 3 weeks I have accidentally learned about long covid / cfs. I have immediately made myself believe I had this and experienced sharp worsening of symptoms - at some point feeling so tired I thought I was going to pass out. I was doing that pacing thing to the extreme, believing I have no energy, for some 2 weeks, my life was hell of only about 1 - h / day of not laying and doing nothing. Not to mention the severe trauma of thinking that my life is over. Trying to do whatever cognitive activity for more than 10 min resulted in a severe headache. After 2 weeks I got covid, I was still laying down all day anyway but I passed it with minor symptoms. 2 days after I thought, wait a minute, maybe it was all fake symptoms, the sharp increase in symptoms after learning about it, the fact that it was not covid that initiated it in the first place. I felt a bit off, slightly a bit off for a few days but I was able to carry on with my days as usual and then I felt 100% well. The problem is that I have got slightly sick one month later and experienced similar, cannot get well, for some one week.Then 2 months later, I got sick again, and it was the same thing with recovering and feeling sick. It's round and round, feeling shitty for 2- 3 days (able to do work though), feeling good for 2 - 3 days, I have had this for a month now but each time it occurs it has a smaller impact. I have yet to find out what it is for sure, I hope it's all made up, whenever it hits I'm actively trying to convince myself that it's not real and try to ignore it. Anyway, I have learned the hard way that hypohondria is not just anxiety / fear about falling ill for some disease, the imaginary symptoms can be much more severe than "my chest hurts, do I have some heart disease?".
I was also a hypochondriac in high school and college, sleeping 12-16 hours a day and still being completely exhausted! Apparently what I really needed was more exercise. The CPAP machine I eventually got after ignoring my PCPs diagnosis merely serves as a placebo, but very effective nonetheless. I don't bother him with my delusions anymore.
I think we can agree that both of those cases aren't worst case scenarios, but rather the only case, or best case scenarios (in the US). The only real way of dealing with health (in the US) is just to never go to a physician and to avoid healthcare at all costs. It will never provide any real answers, despite charging amounts that no individual will ever be able to pay.
The only thing that healthcare in the US really seems to be capable of dealing with is ER, or more specifically clear topically observable physical injuries by mechanical force. We're still not good enough at it to fix things to be as they were before an incident in all cases, but we're closer than almost any other area of medicine.
Your totally generalized dismissal of U.S. healthcare, which despite its many known flaws, routinely treats millions of people in the U.S. per year very effectively, is way the fuck off base and absurd. Not all doctors are the same, not all patients are the same and not all outcomes are worst case scenarios on the part of doctor incompetence. I really hope nobody else takes seriously such a blanket pile of exaggerations to the point of nonsense.
The subtext you're missing here is that GPT with access to the entire corpus of medical data could undermine the entire money printing machine (referring to US healthcare here). What test would cost thousands of dollars if the only human cost to run it is drawing some blood and putting it in a machine?
Abstractions like that are meaningless. Why is the large hadron collider so expensive it’s just throwing stuff at each other and looking at what happens? Random teens could do that…
The actual process of doing blood work is often surprisingly complicated.
Funny enough, I'm actually developing an AI-powered biomedical technological breakthrough that's about to disrupt the medical industry. It uses wearables to enable a blockchain-enabled data management system that also functions as a cloud based SaaS provider, linking with ChatGPT and NFTs to create value for those who are underprivileged, all with only 1 drop of blood. If you want to further this please give me a few billion dollars and I promise something might come :)
No, the variability between individual drops of blood are too high for the precision needed to do most of the tests they were proposing even before you look at the tiny fraction of a drop you would be working with to preform multiple tests. Worse you get excessive contamination between blood and other substances from a finger prick.
So, no very quickly people were calling Theranos either incompetent or fraudulent.
"the tests they were proposing" - yeah, I would agree that some (probably most towards the end) of the tests they were trying to make part of the product were a big part of why they were so fraudulent.
I agree that it was a bad company.
My point is that the general idea was not a bad idea, and because it was so badly mismanaged (trying to add too many products and run before walking, using other companies' tech illegally, etc, etc) we now are worse off because few people want to attempt doing the possible ideas in this space (because people think none of it is possible - which is simply not true).
The company that could reliably be made is far less 'attractive' of course, but it would still be quite beneficial.
Of course... that's absolutely true, and I didn't say otherwise.
However, the basic idea actually is possible, but only for a very select few of the tests they proposed (which they still unfortunately didn't really get right of course).
My point is that, unfortunately, the tests that are possible to do now won't get a company to form out of it - simply because they ruined the scene for everyone. Anyone trying to get a company going in this area and raising money for it will be met with a complete block of resistance because of the precedent set, even if the technology is actually completely reliable.
The last test someone I know had done cost $600 at the physician.
We looked it up and it cost a little less to *buy the FDA approved equipment for your house, and run the test every day for an entire month*.
It's absolutely heinously broken - almost completely due to insurance companies. Insurance should be made illegal and obsolete due to the insane hindrance it is to progress in humanity.
There’s often many tests for the same things with various levels of accuracy and cost.
You really didn’t provide enough information for a more detailed analysis, but for example there’s a reason hospitals don’t just use the cheap COVID tests and call it good.
It wasn't a viral test, and I didn't lay out details here because I don't want to put all my health information on display exactly, but I have several graduate degrees in both engineering and in biology related fields, so I'm quite certain that my analysis of the equipment and technical parameters is correct. It's the financial aspect of healthcare that is horribly wrong, not the science (only let down there is an academic one, as it's awfully basic and antiquated technology that makes it to the clinic, but that's not a bug - it's a feature for safety).
There is a reason I mentioned FDA approved equipment though - it is substantially more expensive to have that label (despite being able to purchase items that claim similar parameters, those need to be exhaustively tested). Despite this, the statement that purchasing some types of equipment for your home and using it regularly, is still often far cheaper than going to the physicians office and getting a test once.
I suppose I am a bit out of the typical demographic though, as I would rather use my own TEM made in my garage to do quick analyses of lab materials than take it to my colleagues across the hall and sign up for a slot on their more resolute EM. Being able to work independently and have your own data is a nice thing sometimes.
I hate doctors and pharmaceutical companies as much as anyone, but those tests are serious business. There's a lot of very hard science and engineering involved in them
When they're cheap outside the USA, it often just means the companies aren't attempting to recoup any r&d costs outside the US
Surely it isn't pharmaceuticals companies doing blood tests in the US? In the UK people's bloods get done in hospital labs (phlebotomy doesn't always occur in hospital though)
Happened to my granma when she was 90. She had convinced her doctor prescribing her 5 or 6 drugs. She swallowed them until she was so weak she had to go to hospital, where the tests revealed that she was 100% sane. She didn't like it.
Then when the patient asks ChatGPT if the tests could give false negatives, it could provide some very valid-sounding answers that say repeat testing might be necessary.
This isn’t a hypothetical. It’s currently happening to an old friend of mine. They won’t let it go because ChatGPT continues to give them the same answers when they ask it the same (leading) questions. At this point he can get ChatGPT to give him any medical answer he wants to hear by rephrasing his questions until ChatGPT tells him what he wants to hear. He’s learned to ask questions like “If someone has symptoms _____ and ____ could they have <rare disease> and if so how would it be treated?” A real doctor would see what’s happening and address the leading questions. ChatGPT just takes it at face value.
The difference between ChatGPT and real doctors is that he can iterate on his answer-shopping a hundred times in one sitting, whereas a doctor is going to see what’s happening and stop the patient.
ChatGPT is an automated confirmation bias machine for hypochondriacs.