Live data from Hacker News

Launching Health in ChatGPT to US Users

openai.com

31–40 of 57 posts

Re: Launching Health in ChatGPT to US Users

#31
I tried posting an Ask HN about my experience using Claude and ChatGPT for medical advice, but it didn't get enough upvotes for any responses. Maybe this is a good place to repost (the part relevant to medical treatment is at the end):

Question: Why don't Claude or ChatGPT get message timestamps?

Go to any conversation you’ve had with Claude or ChatGPT in their chat apps and ask them what time they sent their last message, and you’ll find they don’t see timestamps for any of the messages in your chat history. They might guess or hallucinate an answer but they’ll be wildly off and if you probe further they should admit they don’t see timestamps at all, they only know the current time. I’ve also noticed generally that if you pick up conversations over multiple days, it doesn’t know that days have passed in between messages.

——

Me: How much time has passed since your last message?

Claude: I don't have a way to measure the gap between my messages—I don't have a clock running between turns, and nothing in our conversation timestamps the messages for me. So I can't tell you how much time has passed since my last reply. Could've been seconds, could've been days.

If it's useful, I can check the current time, but that would only tell you now, not the interval since I last wrote.

——

This is very bizarre to me and seems like an obvious flaw in the product, but both major AI companies are doing this and not implementing the obvious solution of surfacing message timestamps. Does anyone know why they don’t do this?

For an example of the type of issue this causes, I’ve been using Claude for advice while my dog is going through medical treatment and it’s very concerning because I’ll say I gave him some medicine and then later ask questions about a symptom he’s displaying and Claude doesn’t seem to know whether the medicine was minutes ago or days ago and usually doesn’t ask for clarification either. I’ve learned to either be explicit about how much time has passed or just not ask it questions where timing would be a factor. But I’m surprised because this seems like it should be an easy thing to address.

Re: Launching Health in ChatGPT to US Users

#33

I tried posting an Ask HN about my experience using Claude and ChatGPT for medical advice, but it didn't get enough upvotes for any responses. Maybe this is a good place to repost (the part relevant to medical treatment is at the end): Question: Why don't Claude or ChatGPT get message timestamps? Go to any conversation you’ve had with Claude or ChatGPT in their chat apps and ask them what time they sent their last me…

I find this weird too! When you say something that prompts them that you’re talking about the current time, they’ll go and fetch it and use that to understand time frames which makes sense. But if you’ve been talking about something tomorrow, and then tomorrow comes and you go back to that chat, they don’t understand that now it’s today/right now.

Re: Launching Health in ChatGPT to US Users

#34

Searching in vain for any mention of HIPAA compliance…

HIPAA only applies to data originating from a covered entity. It doesn't apply to information the patient supplies directly. As a patient, I can take out a street sign displaying my health records if I want. It's my data. If I want to upload it to a strange AI or any other app, that's completely fine. HIPAA isn't involved, for either me or the company I upload to. In fact, quite the opposite. CMS has been going to gr…

It’s a pretty fine line, and if they want to do anything beyond just telling you things about your data, actually taking any sort of actions or communications with health care providers, they should qualify as a business associate under HIPAA. And I certainly am never going to share my healthcare data with an AI company without any sort of regulations or oversight, beyond a TOS.

Re: Launching Health in ChatGPT to US Users

#35

I tried posting an Ask HN about my experience using Claude and ChatGPT for medical advice, but it didn't get enough upvotes for any responses. Maybe this is a good place to repost (the part relevant to medical treatment is at the end): Question: Why don't Claude or ChatGPT get message timestamps? Go to any conversation you’ve had with Claude or ChatGPT in their chat apps and ask them what time they sent their last me…

I'm just guessing here, but I find that the models are often quite bad at ignoring information.

So if the harness appended a timestamp to each message the model sees, I wouldn't be surprised if the model often responded with "given that it's 12:30 pm on a Sunday..." even in cases where that's absolutely not relevant.

Maybe you could instead give the model a tool it could call that would return timestamps?

Re: Launching Health in ChatGPT to US Users

#36
post #8

I think doctors are going to lose their minds or at the very least get extremely agitated/frustrated as patients increasingly use AI tools to do self-research for symptoms and diagnosis. And I know this because I am guilty of this myself and recently had a dentist tell me to not do that. And this applies to so many health workers, your GP as well as your physical therapist, etc. Recently I had Claude analyze an MRI o…

Theres a good use case in AI for patients to chat to tho, if i could transfer all of the context of a conversation I had with a AI bot to my Dr before a appointment it would most likely improve outcomes because I have been much more verbose and descriptive to the AI bot. You basically want an AI chat bot for patient history...

45% of medical note taking software hallucinated details about patients. This is not a good application for AI.

https://www.auditor.on.ca/en/content/specialreports/specialr...

Re: Launching Health in ChatGPT to US Users

#37
post #28

Americans will do everything but not have socialised health care even though they are already leading the world on taxes spent per capita for health care. https://en.wikipedia.org/wiki/List_of_countries_by_total_hea... Also, what is HIPPA?

It’s a patient privacy law that applies to hippos.

Re: Launching Health in ChatGPT to US Users

#38
post #8

I think doctors are going to lose their minds or at the very least get extremely agitated/frustrated as patients increasingly use AI tools to do self-research for symptoms and diagnosis. And I know this because I am guilty of this myself and recently had a dentist tell me to not do that. And this applies to so many health workers, your GP as well as your physical therapist, etc. Recently I had Claude analyze an MRI o…

I like it when patients research things. Consult is more engaging for me when people care about and try to understand stuff. It's hard for patients to interpret medical research and evidence though... ChatGPT recommended treatment for some of my family members based on studies with 60 patients. It wasn't exactly established medical practice.

I was poking around with some symptoms into Gemini not too seriously. It suggested a diagnosis which I have since forgotten. I looked it up and it seems like something with less than 100 known cases in human history.

I think someone needs to teach these models the old "horses not zebras" adage, or we're gonna have a lot of people insisting on having rare cancers.

Re: Launching Health in ChatGPT to US Users

#39
post #7

The American Medical Association and health insurance industries can all fuck off. The healthcare situation in the US is so absurd and Kafka-esque I welcome LLMs in medicine. I haven’t seen a MD in years, I’ve only seen nurse practitioners and physician assistants. Add to that how insurance will randomly deny tests, procedures and treatments, I have very little use for the current medical establishment.

It sounds like you need a better patient advocate or your area is very underserved in provider choice. Many people have success with the health system.

Re: Launching Health in ChatGPT to US Users

#40
Specific tools for MDs to record patient notes were found to hallucinate patient details in 45% of cases. They were especially bad on patient mental health. I am against AI in healthcare in most cases, but anything other than augmentation is probably never going to be happening.

https://www.auditor.on.ca/en/content/specialreports/specialr...

Post reply on HN