A lot of good can be done by extending "one size fits all" solutions for simple things there's just no capacity for
There are so pitifully few health care practitioners compared to what we need in most developed countries I don't think too many jobs are going to be at risk any time soon if AI picks up some of the slack.
That is the artificial scarcity he was referring to.
There is no reason to lump all medicine together in a MD degree /certification.
Why not have one certification for each common problem?
And that is the bottleneck more than any access to medical degrees.
You or anyone can likely go to Caribbean and get an MD. Whether you will finally be able to practice in a field that will make it worth your while is a whole nother story.
90% of medicine is stuff like people going to the doctor with the flu to get a sick note to stay home from work and instructions to rest and drink fluids. I don't think you need an MD for that.
It goes the other way as well. My aunt is an RN and she loves to tell the story about the time that a doctor was in such a rush one time she had to call his attention to the fact that his patient was dead.
It seems to me the US healthcare system sits in this weird gap where it's not funded by private individuals, but it's also not publicly funded (excluding medicare/medicade/VA/tricare etc)
So it doesn't benefit from efficiencies of a single-payer system like the UK has, and it doesn't have to obey proper market forces either, since you have a bunch of for-profit entities in the middle, but the costs are obfuscated from the consumer unless you're unlucky enough not to have employer-funded health insurance.
So until you fix those core issues, I don't see how technological advancements are going to have much of an effect either positive or negative.
It's an orthogonal issue.
For instance, if you read up on the frankly cruel residency process which doctors are forced to clear in the US, the insane hours are in large part because one of the doctors who was influential in shaping the current practice was an abuser of stimulants.
So I think it's a bit of a "just so" story to assume that the version of the medical profession we have is the best possible version.
And indeed Cuba is a good counter-example to your argument. Unless Cubans are just a lot more clever and resilient on average than Americans for example.
I.e. I don't need "Dr House" to tell me to take 2 and call him in the morning. There is a whole large swath of routine medicine which AI could probably do.
And I am happy if you and your family have been lucky to be blessed with good health and good care, but many people have experienced having a mysterious medical condition, and going to several doctors or waiting months before it can be properly diagnosed. Human medicine is far from perfect.
I mean arguably subsidizing med school would be a pretty sensible thing for a government to do if it wanted to increase the number of doctors. The other elements of the venn diagram can't be modified I agree, but affordability of med school doesn't have to be there
1. Go to a GP and explain symptoms 2. GP orders some tests based on symptoms 3. GP refers you to a specialist based on test results 4. Specialist orders more tests 5. Specialist analyzes results and orders more tests, or prescribes treatment plan 6. Check in with specialist in weeks or months to evaluate results and update treatment
It seems to me that for many common issues, 1-4 could plausibly be replaced by AI guided care with minor downsides.
If you are normally spending a lot of time waiting between steps for a practitioner to become available, then it's going to be a net win for many people.
Ironically it's probably easier to replace a lot of what doctors do with AI than what nurses do.
Things must really really bad for someone to defect to Brazil
2. AI can hold vastly more knowledge in its "head" than a doctor. There are many rare conditions that a doctor rarely sees. That doctor doesn't have enough time to spend researching these kinds of things for their patients, but an AI might be able to analyze their medical records/test results/etc. and come up with possible diagnoses that a regular primary care physician couldn't. If these diagnoses go to the PCP for evaluation (as opposed to going straight to the patient), what is the downside or harm?
This is not true. And it's where I believe your whole response takes a wrong turn.
Why is it wrong? Because of risk of harm.
Giving out wrong information can easily kill people. Or permanently insure them. There's a tremendous amount of harm that can occur if you do things wrong in medicine.
When you attach "an AI said X" to something, it adds the appearance of authority. This makes it incredibly dangerous. Folks that don't know enough will believe that the "advice" is true because it "comes from an AI" (and how could an AI be wrong? It's advanced! It's smart! It _can't be wrong_ because the company is using it instead of a real Doctor, and _they wouldn't want to harm me with something that can be wrong_...right?)
For example, let's take COVID. If you had GPT say "go take hydroxychloroquine for COVID", you'd end up with a lot of people getting harmed by (1) taking a drug that's not going to help them when they _think_ it would (so they're not going to do something else for treatment, because they incorrectly think they're already being treated) and (2) there's always side effects to taking _any drug_, and these side effects may actually cause _more harm_ than doing nothing.
What I disagree with is the way that you're drawing the conclusion that there isn't even a valid argument to be made for using AI. To be clear, I didn't say we absolutely should do this - the part of my post you quoted is where I said there's an argument for it (i.e. we should have the discussion and evaluate).
If an AI says take hydroxychloroquine for COVID then absolutely that will cause harm. But that theoretical possibility isn't enough to determine whether an AI should be a source of medical advice - you have to actually evaluate whether/how often it will do that. If .0001% of the time it says to do that and the rest of the time it gives good advice on how to treat COVID at home and what symptoms to look out for that would cause you to need to call an ambulance, then the harms are probably greatly outweighed by the benefits.
This is especially true in the case that I argued - where the alternative is no medical advice, because people don't have access to doctors. Lots of people died at home of COVID when they could have been saved at a hospital - we have to consider what the benefits to them might be in addition to what the detriments of AI giving bad advice could be.
The last thing I'll say is that your position seems to be that if a source of medical advice sometimes gives bad advice, we should not use it. The problem is that based on that position, people should not go to doctors. Doctors are humans who make mistakes - plenty of evidence of that to go around. All of these points that you've made apply to human doctors:
> Giving out wrong information can easily kill people. Or permanently insure them. There's a tremendous amount of harm that can occur if you do things wrong in medicine.
> When you attach "[a doctor] said X" to something, it adds the appearance of authority.
> If you had [a doctor] say "go take hydroxychloroquine for COVID", you'd end up with a lot of people getting harmed
I want to point out that there absolutely were doctors who told patients to take hydroxychloroquine.
This is all to say that I think you have to weigh relative harms and benefits. Should someone with access to top tier medical care get advice from an AI? Almost certainly not. Somebody who has access to no medical care or to the sort of doctor who prescribes hydroxychloroquine? As I said initially, it's not necessarily a hard yes but there is certainly an argument to be made that it's a yes.
I am guessing the latter though I don’t know for sure.
That doesn't mean that there aren't negative influences in modern society. The rich wanting to get richer is definitely one of the major ones that is particularly powerful right now.
But that's why society is a process. You gotta fight back and moderate the negative influences, and boost the more positive ones.
I think the question whether AI will fall on either the positive or negative side of the ledger is wrong in itself. AI will provide a lever to magnify whichever direction society and people were going in anyways independent of AI. It's a tool, and hopefully a really powerful tool, but it's not a God. It can be used equally for good and bad.
And if you believe, like I do (and by all appearances you do too), that our current society is balanced to benefit bad actors over good actors (where bad/good are being defined in terms of benefiting society overall), the magnifying effect of AI is likely to magnify the net negative.
The one advantage of something new like AI coming around is that, I believe, the primary reason open democratic societies allow themselves to become tilted in favor of bad actors is because of complacency...it's usually a slow shift rather than a single big change. The arrival of a major chance like AI forces society and everyone in it to once again take stock of where society stands.
This gives us an opportunity to change the tilt of our society to promote good actors again, before AI truly starts having an impact, which gives us an opportunity to ensure AI magnifies a net benefit, rather than a net negative, for society.
And frankly I think we are closer to having that reckoning than we've ever been in my few decades on this planet, so overall I'm optimistic about AI. But again, whether we benefit or suffer with AI will be independent of AI itself.
You know, I don't think they do think that?
I think they think that's what society should be, though.
If you went completely hungry yesterday, nearly starved to death, but today you get enough calories that you feel less pain so you can at least stand up and walk around a bit, while you watch your brother stuff his face with so much cake he has to puke 3 times to make room for more, to then play frisbee with the rest, what would you think? How it was for you yesterday isn't the standard. How it is for others, today, is the standard. It's about fairness and human dignity, nothing more and nothing less. If everybody was starving, we wouldn't even have a name for it, or a problem with it. It wouldn't be a slight on anyone.
Some people shriek in anguish or sob in darkness, kill themselves, over issues that would be a 2 minute phone call and a check that is 0.0001% of their monthly income for someone else. That's a problem, and that kings used to have less than the average homeless do today doesn't change that, and isn't something capitalism gets to take credit for either. Humans have been inventing and improving things since before history, even animals do it, that's just normal.
I think sociopaths should be called out as the dumb fucking sociopaths they are.
Why the hell should that matter? If the system wanted to keep poor people poor, won't the poverty rate stay the same? Thats the assertion I was countering.
Why keep moving the goalpost? But I will play along:
> If you went completely hungry yesterday, nearly starved to death, but today you get enough calories that you feel less pain so you can at least stand up and walk around a bit, while you watch your brother stuff his face with so much cake he has to puke 3 times to make room for more, to then play frisbee with the rest, what would you think? How it was for you yesterday isn't the standard. How it is for others, today, is the standard. It's about fairness and human dignity, nothing more and nothing less. If everybody was starving, we wouldn't even have a name for it, or a problem with it. It wouldn't be a slight on anyone.
1. You should first ask if your brother got his cake fair and square. If so, then you should be happy you are not puking and be happy for him. Don't be selfish. Let's say your brother worked 10 times more than you (80 hrs) and smarter than you while you stayed in bed and watched TV all day on his dime. Isn't inequality the fairer but still painful outcome in this case? F
2. Should you and your brother have the same quality of life despite the efforts you put in? Why should you be entitled to his fruits of action beyond sustenance?
3. If you agree there in diversity in level of effort and thought, then inequality is the fairer outcome.
4. Also, one should stop comparing only outcomes, you should also compare inputs.
One needs to look at circumstances that allow different inputs in the first place.
Your capacity for input is not some inherent core part of your being, unless you believe in souls, it is a result of what surrounds you and what has come before you.
And you accuse me of moving goalposts? You're not worth my time at any hourly rate.
That's the same with every efficiency improvement. Efficiency improvements (like AI) aren't zero sum. When production costs go down, competition forces prices to also go down, which then leads to more wealth for everyone not just the capital owners (even if it disproportionately benefits capital owners).
People are made redundant and then find jobs in other parts of the economy where they're actually needed. This is a painful but necessary corrective mechanism for the economy to remain dynamic and productive and innovative.
That said, I still want a global wealth tax.
EG you you have 100 people who can't afford steam off a hotdog and 100 customers with n units of disposable income per day. You end up with 150 customers who can't afford shit and 50 with 2n units of disposable income. The price was never bounded by the actual cost in the first place so the smaller wealthier clientele leads to higher prices.
One could easily imagine many tens of millions of new poor making much less while prices driven by wealth of other segments of society continues to rise who both have less coming in to pay higher prices.
Meanwhile you have tens of millions of people who are presently create little value NOW whose jobs end up politically difficult to deprecate. You need to have SOME people on your side after all.
I don't see our society being a good one now or in 20 years. There is no political will to do anything hard and a thread of greed, cowardice, anti intellectualism and weakness woven through our society.
That's how it's always been historically, but you are right that I can't predict it'll be the same in the future, especially when it comes to AGI.
> You are also assuming collusion explicit or implicit and the wealth of remaining buyers doesn't just keep prices high.
Collusion to fix prices can't happen in highly competitive markets with declining production costs, because of game theory reasons. You always get this graph[1] unless you have regulatory capture such as with insulin supply in the US. Or unless you have a labor-heavy cost base such as with the education industry.
[1] https://pbs.twimg.com/media/E314pxwXMAI2ZIn?format=png&name=...
The risk with AI is that it won't be highly competitive. As a general rule, only in a monopoly-like situations can you have abnormal profits. That's where the global wealth tax comes into play, among other policies targeted at monopolistic practices.
The US is particularly bad in this respect. Market sources only seem to prevail in rare cases - like elective procedures which are not covered by insurance (i.e. laser eye surgery).
It's hard to make decisions based on price when you don't even know the price in many cases before receiving care, and you may not have any choice of provider.
It doesn't explain how Sears managed themselves into bankruptcy, handing everything to the likes of Amazon despite having all the money in the world, comparatively.
It also doesn't explain how EVs are taking off. I've seen multiple conspiratorial documentaries about why the oil companies will never allow EVs to ever take hold.
So far, AFAICT, the drive is to cut costs by saving the time the physician spends with the EHR. This has been a goal of the EHR systems for as long as they've existed as it saves money. There is a lot of press releases about products that do more but it's not at all clear to me that these products being widely used.
The UnitedHealthcare example in this article is interesting in that is was being used in production. OTOH, it sounds much more like something evaluating business rules or some heuristics rather than an LLM or even an ML based solution.
[1]: https://blogs.microsoft.com/blog/2023/08/22/microsoft-and-ep...
[2]: https://ehrintelligence.com/news/ehr-clinical-note-length-co...
It's pretty damn impressive tbh.
I'm fairly certain that at least one insurance company VP out there is using the LLM/ML hype to trumpet what is actually a statistical model evaluation. OTOH if it provides cost savings does the business truly care what it is? Especially if they can spin it as 'The computer makes more informed decisions thanks to AI'.
How do you verify that what gets written down really is what the doctor wanted? How do you automate that to protect yourself against LLM hallucination?
In mental health a lot of providers don't even file on customer's insurance. Patient pays cash/charge and the patient is stuck with filing claims. Providers washed their hands of that dirty work.
During my last job (large firm large metropolitan city) employer subcontracted for covid testing erecting a large tent to handle 8 covid exams at once. Here's the procedure:
- you badge in; subcontractor used our hw/network.
- you them give them your name and DOB, which the PC already told them
- they write your name on specimen collection tube, and send you to bay X. The nurse at bay X sees you coming
- at the bay you are asked for your name and dob again. And you badge in again.
That level of duplication can only arise as a function of billing my then employer for extra non productive labor or fraud cya protection. Either way it's damn annoying.
At my last doc appointment I spent the first 20 mins answering THEIR BILLING questions: was my name xyz? Did my insurance change? Etc. No services have been rendered yet.
And despite all the IT for billing and PC displays displaying common sense health tips, there was no display of how long one would wait for the doc. It was a blind date: maybe they'd show; maybe you had better things to do.
In sum, whatever crappy IT health care has, it's to help them, not customers. God damn: they made the paper work hell they live in and, as I say, mental health care patients are stuck with the dirty work because they can't stand it anymore.
My wife spends 2hrs a week yelling at aetena. She's pretty sure they decline claims out of willful stupidity or nefarious greed. I would not be surprised to hear crappy AI involved.
I absolutely hate that business.
I was solicited multiple times for interviews by health insurance corps. I told them NFW. Their pay sucks. Their concept of quality improvement sucks. They suck.
Precious few actually make a positive difference.
Obviously that's not the story they'll tell you, or even themselves, but once you start scratching the surface you'll find that they're all about squeezing profits and pushing costs just like everyone else.
It's probably because you don't understand how the world works. Most things aren't revolutions they just allow us to good things more often and cheaper and so create wide spread "wealth". In this case imagine a doctor who isn't magically better than all doctors combined, but instead is the best of all doctors and specialists and performs very consistently and is available on 24/7 basis. It didn't revolutionize everything but I imagine the collective impact on health and lifespans will be tremendous.
This is exactly what’s happening, it doesn’t even need ai just an algorithm that allows companies to have an excuse for denying care. If there is no human in the loop nobody is responsible. You can’t send an AI to jail so they can just abuse the system.
Judges need to force companies to stop using algorithms and allow approve all expenses when an algorithm is found to be illegal or potentially illegal and that will end the practice quickly. The burden of proof should be on the algorithm writer to show its legal
1. AI doesn't do anything new. It's just human thinking done better in the best case scenario. So the plus side is that we may come up with ideas that a human might have come up with 10-15 years later. However, a lack of ideas is not the problem we face at all. We have solutions to nearly every problem we face. We have enough wealth to implement nearly every solution we have. The real issue we face is that the wealth is concentrated in certain hands that do not want it to make it to others' hands, and those who are capable of implementing the solutions we have prioritize maximizing the wealth they capture with those solutions. AI, as long as it remains in private wealth focused hands, will only serve to turbo charge these issues.
2. Under the current regulatory regime the responsibility of what AI does wrong is not clear. Much like how Uber, and the other gig economy companies leveraged the opportunities created by the spread of the smartphone to create a regulatory innovator, whose only real purpose was to dismantle regulations and impose costs on 3rd parties and capture profits for themselves, AI will and is initially again being used as little more than a legal innovation, using the gray area in terms of legal responsibility as an advantage to do things a company would never have a human employee do or to allow companies to use other people's works without paying for it, and not getting sued for the theft.
Expect AI to continue spreading into more legal gray areas as companies take advantage of the lack of clarity on legal responsibility to raid and pillage our economy and societies.