[nyreed]

Don't see what an LLM really adds here that wouldn't be served by a sensible algorithm which flags up results needing urgent action. Maybe the end user get more clearly explained and non-bewildering information, which is useful, but also points to a failure of any health system / internet resources to provide this.

The LLM _HAS_ empowered the patient to act on his own health, which is undeniably a good thing.

Not sure about seeing the LLM as a companion.

[FeepingCreature]

It seems to me what the LLM adds is that it is the sensible algorithm which flags up results needing urgent action in a growing set of everyday domains.

[djmips]

The human like conversation urging to go to the ER was particularly useful in this instance. I am not saying I don't fear for it's mistakes but clearly it's ability to communciate in natural language is something to not take for granted.

[rasse]

These algorithms do exist.

Indeed the victory for the AI here is the ability to convince the author to take immediate action. The second win for the AI is the concise one liner the author tells the triage nurse so that her further diagnosis and treatment may proceed. A regular human being would be likely to start with a long form story of not feeling well for some time.

[tim333]

It has the benefit of existing and saving the author as opposed to some hypothetical algorithm that doesn't.

[Out_of_Characte]

To be fair, This is what AI is good at. No matter what problem you're trying to solve, AI will provide a solution.

In this case, lab result analysis was entirely based around standard care protocols that LLM is really good at regurgitating back to you.

Yeah sure, anyone could probaly promt feed this into a homeopathy solution which would have killed the patient as per their own predisposed will. You could argue a standard programmatic diagnostic evaluation would be better but its not entirely clear wether human decision making improves based on webMD above an agitated Chatgpt urging you to seek professional help.

And now that I think about it more, no one except doctors and AI could reliably improve diagnosis based on a picture and lab results side by side. So if your healthcare system doesn't provide a response then chatgpt might actually be the closest neighbour in time of need.