@Throwaway05's banner p

Throwaway05


				

				

				
3 followers   follows 0 users  
joined 2023 January 02 15:05:53 UTC

				

User ID: 2034

Throwaway05


				
				
				

				
3 followers   follows 0 users   joined 2023 January 02 15:05:53 UTC

					

No bio...


					

User ID: 2034

This is a good comment - but we are down some staff at work and I know by the time I swim out of this it will be hard to motivate myself to circle back, I didn't want to leave you totally hanging though.

Although one specific thing, a common problem is people having a viral illness and demanding antibiotics (which at best would do nothing). Anti-virals generally being useless is a separate kettle of fish.

Doctors do that all the time, especially if there are adverse financial incentives at play. The US is going through a multi-decade crisis because of opioid overprescription and people shop around for ADHD stimulants.

I hope you can see the difference between a structural problem that consists of some bad apples and institutional incentives, and a process that delivers glazing 100% of the time unless you take significant active steps to avoid it.

Thanks!

Do you have a source for that?

You have to keep in mind that a relatively small fraction of the population (often with relevant diagnosis such as personality disorders, anxiety, illness anxiety disorder) are super utilizers of services, inappropriate users of services (such as misuse of mychart physician messaging) and are disproportionately the people complaining online (as well as misusing AI tools).

You see another version of this in the anti-psychiatry movement where people who can't recognize they have an illness because of the illness (ex: schizophrenia) are complaining about being forced into the hospital and take medication when that is the correct thing to do.

As for rationing the simplest example of this is as a good example of over utilization that many people fall into - suffering sometimes is part of life. If you have a viral illness you need to wait it out and take care of yourself most of the time, the doctor will not be able to do anything and rarely is it worth confirming which URI it is. Antibiotics will not do anything useful.

This does not stop people from emphatically making demands.

Yeah that's fair, we definitely have many posters here who would use a tool like that and be sensible about.

We also have even more people who would do similar stuff and be intensely foolish, and have no awareness of that fact.

Unfortunately the productivity gains from the former are probably outweighed by the damages from the latter.

If AI was able to write prescriptions for me, I would choose Astra over 95% of doctors.

If you really believe this you should consider your biases. A response to a post noting that AI is telling patients to eat TP should not be "more of this please!"

Especially since the LLMs are well known to flatter the biases of their users, which is something doctors are changed not to do due to the negative outcomes associated with that.

The problem is that patients do this now and it's a disaster and they don't realize it. Everyone who does outpatient medicine is flooded with instances of patients being sure about something and it being wildly off base, useless, and time consuming because the patient has been convinced of some nonsense that we have to walk back. I'm sure some fraction of people are saving themselves time or getting useful information but it's not well represented in what we actually see.

Part of this is because people don't realize that a lot of medical work involves things like dealing with an anxious patient who repeatedly asks the AI until they get the answer they want, assholes, and things like rationing care (which the AI has no facility with).

Believe it or not a lot of visits (especially to the ED) are pointless and we'd love it if people were successfully using AI for triage.

They aren't.

I think in a just world you might be right but the reality of the situation is that things have become wrapped up in comorbid political problems and a focus on say, jailing Fauci just firmly distracts from getting any useful done.

You are never going to know who punched you in the face. You can take steps to avoid getting punched in the face again, regardless of who did it, or you can obsess over who punched you in the face and get punched in the face again and again.

I'd love to know who punched you in the face. I'd love for you to never get punched in the face again, and I'd love to never get punched in the face.

No amount of pressure from the West or anyone else is going to make China stop being sketchy as fuck, it's unlikely we will ever know.

Improving wet market and lab safety is an increasing existential threat.

Don't focus on the former at the expense of the latter.

Just as public health authorities did themselves no favors by being misleading and overly restricting, COVID skeptics do themselves no favors by overstating the case - both explanations make sense and are plausible, and any conclusive evidence is likely gone. Focus on patching the vulnerabilities over assigning blame and asserting informational superiority.

nd various schemes which are basically "AIs checking other AIs,"

This was my first thought, has this gone anywhere?

The problem is the error rate (for professional use) and misuse with accompanying over certainty (for amateur use).

Great Another AI Post.

On the front page of Meddit today is a post that basically goes "AI convinced my patient that eating toilet paper helps with constipation." This to my experience appears to be the norm.

In Medicine I've found AI to be of limited utility, because mistakes are not allowed and the required error checking usually nullifies productivity gains.

In Programming it seems to be hot shit because error checking is part of the job and required either way.

Other fields things get more complicated and I have less experience. But the usual people seem excited and the usual people seem frustrated.

So I want to know, for those of you who use AI for work outside of programming (and not to simplify tasks for hobbies/personal life) what are you actually using it for, how is it helping your discipline? I am endlessly fascinated by the immense gap between people who love it and hate it.

Also, for those of you who know a bit more about LLMs, the lack of true error correction seems to be a fundamental problem, is their any hope of this actually being addressed without completely changing away from LLMs?