Hacker News Reader

I’m a co-founder at a YC startup (https://empirical.health) that treats a number of long COVID patients, and I think the way forward really is to integrate AI into the doctor’s work.

GPT-4 is often scarily good at suggesting diagnostic and treatment plans for even very rare conditions. It will sometimes suggest things even well-trained doctors wouldn’t think of.

But it also makes some basic and expensive errors. For example, it often suggests long COVID patients with tachycardia get an echocardiogram (which is unnecessary and can cost hundreds of dollars).

So I think the model where AI drafts, but a doctor edits, would lead to much better patient care.

You are using ChatGPT correctly, it is great at providing reasonably accurate background explanations on terms and concepts in complex fields. You are using it like someone would use wikipedia or a reference book, but saving yourself many trips to the library or having to search through a lot of different textbooks. Of course it's best at this for concepts that you would expect to be well documented in it's training set.

Thank you, that's exactly how I view it!

For what it’s worth my dad was in the hospital for 3 weeks with Covid March 2020 two of which he was in a coma. Everyone in my family had symptoms but me, months later I developed strange auto immune symptoms like nerve pain in my face and feet.

Only thing that I found that helped was a strict carnivore diet and weight lifting (med-heavy) for a month and all of the symptoms subsided. I put more stock in the weight lifting, YMMV.

So weird, stories like this have convinced me that nutrition/diet is deeply person dependent and it's all about finding what works for you personally instead of some global optimum

just keep in mind there are other options which do not have an official policy which claims using your chat logs to “develop models” is “illegal, harmful, or abusive” (a policy which ensures your learning from using that LLM cannot be included in larger datasets to help yourself or others in the future without uncertain legal risk many people and organizations will refuse to accept)

Gemini and Llama have no such customer noncompete.

I have to assume that clause addresses people directly training new AIs on ChatGPT conversations, not what you do with personal knowledge gained from using the tool.

Training new models on ChatGPT conversations was a pretty common way to "cheese" getting local models up to snuff (or was, before they got there on their own)

I can't imagine it somehow binds you to never redistribute the knowledge you now have in your head, thats literally the same argument these AI companies have been making as to why its not copyright infringement when their model scoops up the IP of everything on the internet and regurgitates the amalgamation.

Gemini has. It is written in Gemini API Additional Terms that you cannot develop a competition model by reusing the outputs of Gemini.

> Doctors in most places don't seem to be taking this condition seriously and believe it to be a made up thing, but COVID causing damage to bodies and brains is a very real thing.

What happened with "trust the experts"?

The experts in this case are researchers trying to figure out new treatments, and ChatGPT helps me understand their research and advice. Most general doctors are not experts on complex chronic diseases that impact multiple parts of the body in different ways. If I had doctors accessible to me who were experts in this then I wouldn't have to rely on ChatGPT, but this is all I have.

Fair enough.

I am just personally wary of diagnosing myself - what I may think of as Long Covid (because it correlates to symptoms I've read about) may just as well be something else. I have no way of knowing.

But everyone's circumstances are individual, I suppose. Especially when regular doctors are unable to help you. That's a tricky situation.

I don’t think someone should use GPT this way but this snark isn’t productive.

It was always lazy shorthand for “we don’t fully understand what’s going on but in general you’re probably better off by trusting the current scientific consensus,” which is a consistently good approach across time and space even if the consensus turns out to be wrong.

And for cases where there isn’t broad consensus, like the etiology of Long COVID, then obviously “trust the experts” is an almost semantically meaningless statement anyway. This is not analogous to climate change or vaccines or other areas where you hear this (again, lazy) phrase a lot.

There have been many scenarios in which there wasn't broad concensus, but the lazy phrase "trust the experts" has been used. Ironically, COVID vaccines when they were first developed were one of those scenarios.

In lieu of trusting experts, what would you propose to have done in that situation? Get everyone to conduct their own double blind trials? Apportion equal weight to the arguments of people arguing the vaccines contained microchips or autism?

I am not proposing anything. I am simply pointing out the hypocricy of trusting the experts when it fits you, and discarding them when it doesn't.

Frankly whilst ChatGPT isn't something I'd trust myself because it's a just big dumb model that fills in gaps with statistically optimal continuations that can be completely untrue, if it works, it works by summarising the consensus of the experts (with controversial perspectives heavily RLHF-ed out, especially around COVID...) with broader and deeper specific knowledge than general medical practitioners offering fewer suggestions, so I'm not convinced consulting it reflects a desire to discard expert consensus...

Not really. There was a randomized controlled trial of more than thirty thousand people backing the emergency approval of the Moderna vaccine.

The results were obscenely strong.

So no, not really. RCTs of that scale and with those strong of results build consensus pretty quickly, which is why there was (and remains) such strong consensus, which is why you feel like "trust the experts" was such a cudgel.

These arguments are always fun because they resolve to, "I'm mad because I wasn't trusted with subtle arguments and statistical nuance" immediately followed by a demonstrated inability to grasp exactly those things. Of course if you really wanted the subtle arguments and statistical nuance, you could've looked at the study results themselves. But you almost certainly didn't and most people wouldn't, ergo "trust the experts" who do those things is often very reasonable advice.

Capitalism happened. Experts are profit-oriented now.

My experience of LLMs as a developer with code is they are excellent at producing reasonable looking stuff that is slightly wrong in hard to spot ways.

The idea of people using them for medicine is frankly terrifying to me.

Hey guess what, it's terrifying to me too. But I don't seem to have a better option at the moment, as elaborated on above.

Yeah, to be clear that wasn't blaming you, clearly the core issue is the lack of real healthcare options.

> The idea of people using them for medicine is frankly terrifying to me.

This reminds me of the self driving car debate.

Plenty of medical people are wrong in hard-to-spot ways. Plenty are wrong in easy to spot ways too.

If we get a machine to do the job, should it be perfect, or should it be better than the average human?

That's a separate issue. LLMs are not better than average software developers, and I sincerely doubt they are better than the average doctor.

I'm not expecting perfect, I'm expecting not literally every non-trivial response to be subtly wrong, which is my experience with code from them. If every time your self-driving car goes out it runs someone over, no, we shouldn't allow it.