Im a doctor, i use LLMs a lot. Theyre great and generally better than the average doctor and will only get better. 90% of people on earth have pretty limited medical access. LLMs can give everyone access to at least great information and diagnosis, and hopefully in the future robots can give everyone access to to surgery and procedures too.
ryandrake · 5h ago
Seems like "Caveat emptor" has become the default regulatory framework for AI. Getting medical advice from an chatbot (let alone legal advice or even financial advice) seems like a very bad idea, but none of these companies seem to want to admit it, educators don't seem to be in a hurry to educate the public about it, and regulators are pretty much entirely asleep at the wheel.
cyanydeez · 3h ago
"Caveat emptor" has become the golden rule for the grift economy. This is definitely a broader late stage capitalism thing, and not just AI.
SpicyUme · 1h ago
Man that's grim. But it seems to be the way things are going right now.
01HNNWZ0MV43FF · 6h ago
We can't afford real doctors
JohnFen · 5h ago
Taking medical advice from genAI can be much costlier.
jpc0 · 3h ago
Using it as a tool isn’t, maybe not everyone would use it correctly but I’m happy to share.
About a month ago I came down with what I thought was a mild fever, usually I would likely not have cared much, went to work and pushed on but I asked the AI anyway. It asked some follow up questions and then recommended I see my family medicine doctor urgently.
A few hours later when I saw my doctor my fever had spiked to above 40c and I needed immediate medical intervention.
Without the AI I might have toughed it out, went to work, went home early because I felt bad and tried to push through to the next day, hoping I could sleep it off.
Instead of a cheap visit to the doctor and some meds and in office treatment, that situation could have gotten way more out of control likely a hospital trip if not worse.
The subscription definitely saved me some money if not more, not because it dispensed medical advice, but because it was available and had enough “knowledge” to recognise an emergent situation.
JohnFen · 1h ago
> maybe not everyone would use it correctly
Guaranteed that not everyone will use it correctly. People already commonly don't use it correctly for less sensitive things. I didn't mean to say it's not useful at all.
The issue is going to be that many people will take what it has to say as authoritative and just follow the advice. Sometimes that advice will be actively harmful. I'm not sure how people are going to be able to tell when the advice should be ignored or not, which, in my view, makes it too risky to use for this purpose -- especially if it's right more often than it's wrong.
About a month ago I came down with what I thought was a mild fever, usually I would likely not have cared much, went to work and pushed on but I asked the AI anyway. It asked some follow up questions and then recommended I see my family medicine doctor urgently.
A few hours later when I saw my doctor my fever had spiked to above 40c and I needed immediate medical intervention.
Without the AI I might have toughed it out, went to work, went home early because I felt bad and tried to push through to the next day, hoping I could sleep it off.
Instead of a cheap visit to the doctor and some meds and in office treatment, that situation could have gotten way more out of control likely a hospital trip if not worse.
The subscription definitely saved me some money if not more, not because it dispensed medical advice, but because it was available and had enough “knowledge” to recognise an emergent situation.
Guaranteed that not everyone will use it correctly. People already commonly don't use it correctly for less sensitive things. I didn't mean to say it's not useful at all.
The issue is going to be that many people will take what it has to say as authoritative and just follow the advice. Sometimes that advice will be actively harmful. I'm not sure how people are going to be able to tell when the advice should be ignored or not, which, in my view, makes it too risky to use for this purpose -- especially if it's right more often than it's wrong.