241–250 of 677 posts
Another nail in the coffin for apps that depend on AI APIs because the AI companies themselves are working on products using their own APIs (unless you can make the UX significantly better). UX now seems like the prime motivator when building apps.
My uncle had an issue with his balance and slurred speech. Doctors claimed dementia and sent him home. It kept becoming worse and worse. Then one day I entered the symptoms in ChatGPT (or was it Gemini?) and asked it for the top 3 hypotheses. The first one was related to dementia. The second was something else (I forget the long name). I took all 3 to his primary care doc who had kept ignoring the problem, and asked…
So many doctors never bothered to conduct any tests. Many said it's in my head. Some told me to just exercise. I tried general doctors, specialists. At some point, I was so desperate that I went to homeopathy route.
15 years wasted. Why did it take 15 years for the current system?
I'd bet that if I had ChatGPT earlier, it could have helped me in figuring out the issue much faster. When you're sick, you don't give a damn who might have your health data. You just want to get better.
Earlier quoted context omitted.
It can also help you in not getting a job because your health data says you'll be sick in 6 months.
It would be absolutely amazing if any sort of tech could say that I'm going to have a serious health problem 6 months ahead of time.
I’m kind of torn on this. From one side, I can’t seem to trust doctors any more. I recently had a tooth removed (by the advice of two different doctors), in a claim that it will resolve my pain, which it did not, and now 3 different doctors don’t know what’s causing my pain. Most doctor advice boil down to drink some water and take a painkiller, while glancing for 15 seconds at my medical history before they dedicate…
- Humans are self healing - if a doctor does absolutely nothing, most issues will resolve on their own or, if they're chronic (e.g. back pain) they won't be worse off compared to the alternative. I'm in a country with subsidised health care, and people go to the doctor immediately for absolutely anything. A doctor could have a 99% success record by handing out placebos.
- Most patients have common issues. I.e. maybe 30 people visit the clinic on a given day, it's possible that all 30 of them have come because they have the flu. Doctors are human, nobody's going to investigate potential pneumonia 30 times a day every day for 6 months. So doctors don't: someone comes in and is coughing, they say it's flu, on to the next patient. If the person really has pneumonia, they'll come back when it gets worse.
- Clinics are overbooked. I don't know if it's licensing, GDP, artificial scarcity, cost regulations or what, but doctors probably don't actually have time to investigate anyways.
- Doctors don't receive any rigorous continuing education. I'm sure there's some restrictions, but I've gone into doctors in the last year and gotten the "stress causes ulcers" explanation for what turned out to be food sensitivity issues (there was no visible ulcer mind you, so it was concluded that it was an invisible ulcer). Slow, gradual maintenance, and heavy reading are hard things that humans are necessarily bad at.
- Patients don't want to hear the truth. Lifestyle changes, the fact that nothing can be done, there's no pills to cure you, etc. Even if doctors could give a proper diagnosis, it could end up being bad PR so doctors are conditioned away from it.
- Doctors don't follow up - they get absolutely no feedback whether most of their treatments actually work. Patients also don't come back when their issue is resolved, but even if they do doctors don't care. Skin issue, doctor prescribed steroidal cream, redness disappeared, doctor declared I was cured, redness came back worse a week later. As a scientific field, there's no excuse for anything but evidence based medicine, but I haven't seen a single doctor even make an attempt to improve things statistically.
I've heard things like, doing tests for each patient would be prohibitively expensive (yes, but it should at least be an option and patients can pay for it) or the amount of things medicine can actually cure today is very small so the ROI would be low for additional work (yes, but in the long term the information could result in furthering research).
I think these are obvious and unavoidable issues (at least with the current system), but at the same time if a doctor who ostensibly became a doctor out of a desire to help people willingly supports this system I think they share some of the blame.
I don't trust AI. Part of me goes, well what if the AI suddenly demands I have some crazy dental surgery? And then I go, wait, the last dentist I went to said I need some crazy dental surgery. That none of the other 3 dentists I went to after that even mentioned. And as you said an AI will at least consider more info...
So I do support this as well. I'd like to have an AI do a proper diagnosis, then maybe a human rubber stamp it or handle escalation if I think there's something wrong...
Earlier quoted context omitted.
Glad to hear your uncle improved! Would you mind sharing the other two hypotheses and what the diagnosis ultimately was?
The first was "dementia" (or something related to it, I don't remember the exact medical term). The second was something to do with fluid in some spinal column (I am sorry once again, I do not remember the medical term; they operated on him to drain it, which is why I remember it). I don't remember the third one, unfortunately.
On second thought — the opposite. A bulge/blockage of CSF?
Here’s something: my chatGPT quietly assumed I had ADHD for around 9 months, up until October 2025. I don’t suffer from ADHD. I only found out through an answer that began “As you have ADHD..” I had it stop right there, and asked it to tell me exactly where it got this information; the date, the title of the chat, the exact moment it took this data on as an attribute of mine. It was unable to specify any of it, aside…
Earlier quoted context omitted.
Depends on the product itself. For example I use an LLM for tracking calorie data by telling it or providing a picture of the food I had, and it does a web search to find those details. The problem is it doesn't actually remember past meals, so I wrapped this API call in a simple app that just tracks the calorie numbers like a traditional app. Just having an LLM is not the right UX for the vast majority of apps.
Right. But what if you dropped the human-facing UI, and instead exposed the backend (i.e. a database + CRUD API + heavy domain flavoring) to LLM as a tool? Suddenly you not only get a more reliable recognition (you're more likely to eat something that you've eaten before than completely new), but also the LLM can use this data to inform answers to other topics (e.g. diet recommendations, restaurant recommendations),…
A solution could be, can the AI generate the UI then on the fly? That's the premise of generative UI, which has been floating around even on HN. Of course the issue with it is every user will get different UIs, maybe even in the same session. Imagine the placement of a button changing every time you use an app. And thus we are back to the original concept, a UX driven app that uses AI and LLMs as informational tools that can access other resources.
Here’s something: my chatGPT quietly assumed I had ADHD for around 9 months, up until October 2025. I don’t suffer from ADHD. I only found out through an answer that began “As you have ADHD..” I had it stop right there, and asked it to tell me exactly where it got this information; the date, the title of the chat, the exact moment it took this data on as an attribute of mine. It was unable to specify any of it, aside…
Disable memories so each chat is independent. If you want chats to shared info, then use a project.
Here’s something: my chatGPT quietly assumed I had ADHD for around 9 months, up until October 2025. I don’t suffer from ADHD. I only found out through an answer that began “As you have ADHD..” I had it stop right there, and asked it to tell me exactly where it got this information; the date, the title of the chat, the exact moment it took this data on as an attribute of mine. It was unable to specify any of it, aside…
[flagged]
My cleaning lady's daughter had trouble with her ear. ChatGPT suggested injecting some oil into it. She did and it became a huge problem, so that she had to go to the hospital.
I'm sure ChatGPT can be great, but take it with a huge grain of salt.