BOTS

[kazzafir]

I am amazed that the forum is staying that a BOT is more emphatic and helpful than a doctor. It's not empathic, it's a robot programmed with responses!!!!! How can people rely on something that is not real?

 

[KennyA]

Hi kazzafir and congratulations on your first post.

As you may know we have a rule against posting AI generated material on this forum. Grateful if you can point me to where you think someone has promoted AI material, please.

 

[JoKalsbeek]

I am amazed that the forum is staying that a BOT is more emphatic and helpful than a doctor. It's not empathic, it's a robot programmed with responses!!!!! How can people rely on something that is not real?

Half the time, people don't even realise they're dealing with a bot... And they're not supposed to be on the forum. Flesh and bloods only, really.

 

[Rokaab]

I'm its assuming its the diabetes.co.uk email that went out this morning where its talking about research on Chat GPT
ie its not from the actual forums, but research done

 

[Oldvatr]

If it replaces those website customer support bots then it will be an improvement. I find them totally useless and I end up saying You are Fired.

 

[KennyA]

@Rokaab - thanks.The forum has no influence over what appears on the website.

Here's a link to the original study. https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2804309

I have a couple of concerns about how good this study actually is.

Firstly, it looked at only 195 responses.
Secondly, these were responses to selected questions posted by unknown individuals to a social media forum on Reddit, and answered by clinicians on that forum, which is not typical of a normal patient/clinician interaction.
Thirdly, the "clinicians responding" on the Reddit forum are unlikely to be giving responses typical of a normal clinician/patient interaction - no physical exam, no access to previous history etc. They should be constrained in their responses and advice by standard medical ethics.
Fourthly, the "benefit" was retrospectively judged by three "licensed healthcare professionals" and not by the person asking the question/receiving information.
Fifthly, the evaluators may have preferred the ChatGPT responses simply because they were invariably much longer - this is not controlled for.
Sixthly, there is the question of whether those carrying out the study have any financial or other interest in the provision of AI services. I can't see the full text so do not know what was reported.

 

[Robbity]

I am amazed that the forum is staying that a BOT is more emphatic and helpful than a doctor. It's not empathic, it's a robot programmed with responses!!!!! How can people rely on something that is not real?

???

See https://www.diabetes.co.uk/forum/th...al-intelligence-chatbots.196357/#post-2603143

IMO there's a very big difference between reporting and recommendation.

But certainly in general AI bots can only be as "intelligent" as those who input their particular choice of information and/or data to them.

 

[Oldvatr]

@Rokaab - thanks.The forum has no influence over what appears on the website.

Here's a link to the original study. https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2804309

I have a couple of concerns about how good this study actually is.

Firstly, it looked at only 195 responses.
Secondly, these were responses to selected questions posted by unknown individuals to a social media forum on Reddit, and answered by clinicians on that forum, which is not typical of a normal patient/clinician interaction.
Thirdly, the "clinicians responding" on the Reddit forum are unlikely to be giving responses typical of a normal clinician/patient interaction - no physical exam, no access to previous history etc. They should be constrained in their responses and advice by standard medical ethics.
Fourthly, the "benefit" was retrospectively judged by three "licensed healthcare professionals" and not by the person asking the question/receiving information.
Fifthly, the evaluators may have preferred the ChatGPT responses simply because they were invariably much longer - this is not controlled for.
Sixthly, there is the question of whether those carrying out the study have any financial or other interest in the provision of AI services. I can't see the full text so do not know what was reported.

Here is another editorial about this study

Chatbot medical responses get good reviews in UCSD study

ChatGPT often answered routine questions more completely and with more empathy than busy human doctors, according to evaluators.

www.lajollalight.com

And something about the institute doing the study. I note that 3 of the prime authors of the study are computer engineers, not medics per se.

La Jolla Institute for Immunology - Wikipedia

en.wikipedia.org

 

[LokiTheFiddler]

@Rokaab - thanks.The forum has no influence over what appears on the website.

Here's a link to the original study. https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2804309

I have a couple of concerns about how good this study actually is.

Firstly, it looked at only 195 responses.
Secondly, these were responses to selected questions posted by unknown individuals to a social media forum on Reddit, and answered by clinicians on that forum, which is not typical of a normal patient/clinician interaction.
Thirdly, the "clinicians responding" on the Reddit forum are unlikely to be giving responses typical of a normal clinician/patient interaction - no physical exam, no access to previous history etc. They should be constrained in their responses and advice by standard medical ethics.
Fourthly, the "benefit" was retrospectively judged by three "licensed healthcare professionals" and not by the person asking the question/receiving information.
Fifthly, the evaluators may have preferred the ChatGPT responses simply because they were invariably much longer - this is not controlled for.
Sixthly, there is the question of whether those carrying out the study have any financial or other interest in the provision of AI services. I can't see the full text so do not know what was reported.

I think most people will recognise something from the article..
its been a long time since people actually got to see a GP , let alone talk to one.
Nowadays you get to talk to a prescribing nurse who hasn't read any of your records and has her own opinions..
I'm on semaglutide/GLP1 and digestion has slowed so as to increase my blood sugars. A lower dose initially helped me not eat, lose weight and have lower HA1Bc. Its gone up since the dose was raised.
Was told by nurse that was due to my diabetes not the drug.
The nurse clearly didn't have a clue about semaglutide /GLP1.
Asked to speak to a GP and just get same response, GPs are only interested in 'new issues..

So its not that hard for a bot to out perform a lot of GPs.
The internet has often been the first port of call for most people then 111, and probably A&E after that if its serious.
Just my opinion/experience.

The NHS don't use the data they collect/have very well either.

I recently had a pill cam video done. took over a month for results :-(
I asked if the video/info was available.. and stored?
Reply: No!
Did they do any analysis of the video?
Reply: A 'doctor/medical professional ' looked it and gave 1-2 lines of comments.
Asked if the peristaltic wave was visible and if they could time it from the video..
Reply: Why would they want to do that ?!

Have sent in a request for the video via email, but I'm guessing its not available. That was a few months ago..

 

[Oldvatr]

I think most people will recognise something from the article..
its been a long time since people actually got to see a GP , let alone talk to one.
Nowadays you get to talk to a prescribing nurse who hasn't read any of your records and has her own opinions..
I'm on semaglutide/GLP1 and digestion has slowed so as to increase my blood sugars. A lower dose initially helped me not eat, lose weight and have lower HA1Bc. Its gone up since the dose was raised.
Was told by nurse that was due to my diabetes not the drug.
The nurse clearly didn't have a clue about semaglutide /GLP1.
Asked to speak to a GP and just get same response, GPs are only interested in 'new issues..

So its not that hard for a bot to out perform a lot of GPs.
The internet has often been the first port of call for most people then 111, and probably A&E after that if its serious.
Just my opinion/experience.

The NHS don't use the data they collect/have very well either.

I recently had a pill cam video done. took over a month for results :-(
I asked if the video/info was available.. and stored?
Reply: No!
Did they do any analysis of the video?
Reply: A 'doctor/medical professional ' looked it and gave 1-2 lines of comments.
Asked if the peristaltic wave was visible and if they could time it from the video..
Reply: Why would they want to do that ?!

Have sent in a request for the video via email, but I'm guessing its not available. That was a few months ago..

pillcam video? Thats a novelty! I presume you were tracking food progress rate through the gut? was the camera in the pill. or was it just tracking a marker using ultrasound? swallowing a camera sounds painful.