@parismarx From a business perspective, the biggest thing about LLMs is not what they can actually do but how they can wash away responsibility. That's a *huge* deal to the rentier class, and "AI" proponents are well aware of it.

@parismarx re: slimy #UnitedHealth automated AI healthcare denials, people need to know that #MedicareAdvantage <> #Medicare! #MedicareDisadvantage

P.S. Mastodon should have quote toots

@parismarx this tracks.

@parismarx it's like corporations are bad and every tool they use is used in the worst way possibly because they're inherently corrupt.

@parismarx AI "death panels".

@parismarx Correct. Skynet is never going to happen. But humans using AI nefariously against each other, that's the threat.

@parismarx And that is the insurer I'm stuck with, for reasons.

@parismarx The EU is putting together some good legislation on this in the AI Act - along the lines of GDPR, so you can't be subject to arbitrary automated decision making.

@parismarx

Stay with regular Medicare. AVOID health insurance companies whenever possible.

@parismarx solution, don't give them your money in the first place.

@parismarx I know, it's a nightmare. Me and my coworkers spend half of our time battling insurance companies.

@parismarx both are still risks, just one is years off and the other starting to happen now.

@parismarx I was assured that the only issue with AI was that it would simulate copies of people and torture them for not working sufficiently hard to bring the AI into being.

@parismarx Request the reason and signature from the doctor advising the insurer deny care.

@parismarx That's more like just plain fraud than it is "AI".

@parismarx i feel like we could have seen this exact headline 10 years ago, but it would have said 'algorithm' rather than 'AI model', probably to describe the exact same software

@parismarx AI decisions aren't really decisions ..they are averaged guesses weighted in the companies favour.

@parismarx Thanks for the link.

I know you're very familiar with it, but I think everyone in your replies with "capitalism is to blame" or "what about misinformation" comments needs to study the Robodebt case in some detail.

@parismarx but they don't need AI for that. It's just something to hide behind, to pretend that denying care isn't simply their profitability model.

@parismarx well the company was going to do the same. Soon they will have AI that tries to guilt trip you into going to work when you are sick, AI that fires you for being queer and AI that drops bombs on villages

@parismarx Imagine if they used it to optimize and streamline appointments, care, and employees were facilitating access. It's like we're so close to getting it right but a few people in suits say that we have to do it their way.

@parismarx "We don't need to pay - we just need to prove legally why we aren't required to pay."

@parismarx Ah ah ah AHEM:
https://kolektiva.social/@LeftistLawyer/111433031244950311

@parismarx

It doesn't matter how accurate it is, so long as it gives the answer that's more profitable. And because it's a black box to everyone who doesn't understand AI, it's a very useful tool for profiting.

@parismarx I hesitate to call this stuff "AI" but algorithmic denials have been an issue for a while. In particular I remember reading something a while back that one (about drug seeking, IIRC SAID it was "for advisory use only" or something like that, but people were taking this potential flag and running with it as if it was gospel... unfortunately I think part of it has to do with liability. "This says there's a chance, so if I disregard this, I could be liable" ...which is shitty as hell.

@parismarx now I have two things I hate: people calling it AI, and people using machine learning to make life all the more hellish.

@parismarx I'm surprised AI is being used for serious cases already when we still lack understanding of many things about it. Assuming this is by design is highly disturbing also.
At the same time, I wonder what is the error rate for handling such cases by a human being.

@parismarx @gwynnion

My old insurer has a human setup to deny you care 100% of the time.

@parismarx Obviously these companies are using AI for nefarious reasons, but this doesn't inherently mean all AI is bad. What we have here is a bad actor abusing AI with a model that's been specifically designed to be biased and deny any and all claims, all to penny pinch those who signed up to their insurance policy and to avoid having to pay out any claims to their clients. It's just fraud.

@parismarx that is in the nature of health insurers. They will misuse any tool in the book to deny people their rightful payments. Nothing has changed.

@parismarx hoping some version of the Algorithmic Accountability Act makes its way into law, at the very LEAST
https://posts.bcavello.com/why-im-still-hyped-about-the-algorithmic-accountability-act/

@parismarx tbf.... United healthcare was already denying a shit ton of claims without AI. That's a horrible company

@parismarx This was almost certainly not "AI" in any real sense. The company is just using the word as a barrier to stop people asking questions. What data was this model based on? Who decided the rules it should follow?

My guess is it was a simple linear regression based on previous insurance decisions, which of course are not biased in any way at all.

@parismarx

You mean that in 9 out of 10 cases the UnitedHealth AI awards funding the patient doesn't actually merit?

...

Thought not.

@parismarx @elecharny to be fair, though, they don’t really need AI to deny claims, though, perhaps they think using an algorithm somehow absolves them?

@parismarx @ljs AI isn't the issue here. Having a for-profit medical insurer is. Using the worst tool for the job that increases profits is only a byproduct of that.

@parismarx if corporations are people this is basically theft and probably some bodily harm in there, why isn’t this corporation in jail, you know since it’s a person

@parismarx
I wish it wasn't called AI, there's nothing intelligent about it.

@parismarx but ML models have already been doing this for years.

@parismarx
Wat #Rutte, #Kamp en trawanten deden/doen met #algoritmes, doen ze nu met #AI ook.

Weet wat je kiest

#verkiezingen2023

@parismarx looking at some of these responses, people are buying the flawed excuse right out of the gate. Any excuse to deny is the intent, in the design. How about AI being used to successfully prosecute the PEOPLE implementing this. What is wrong with us!!! to accept this for a minute, pay for a service you'll never get. This is more than never getting a meal you paid for or cable that never works. The consequences of this embezzlement are lethal for some; others suffer afflictions with no

@parismarx
"Health" Insurance companies use an AI model with 90% error rate to deny care.

Meanwhile, the big insurance companies are running an ad on TV featuring @BernieSanders fearmongering over a "government takeover of your healthcare!"

I WISH the gov't would take over my healthcare! https://www.washingtonexaminer.com/news/senate/conservative-watchdog-ads-against-bernie-sanders-pbms

@parismarx Bad vibes for UHC. AI, such promise, such danger. 🤨

@parismarx While 100% fcked up, I would think that the crux of the message here should be that in 2023 it's still possible to deny needed healthcare.

@parismarx I think the 10% rate is enough for them. saves money

@parismarx to the insurance company the errors are a feature not a bug.

@parismarx The problem with the “fantasies” is that when their feasibility will become evident to a non-specialist it would be waaaaaay too late to do anything.

Meanwhile, enjoy fighting your corporate overlords. There’s no real need to dunk on other people working on other problems.

@parismarx

👆​
#MurderForProfit
#UnitedHealthcare
#NationalizeHealthcare now
#MedicareAdvantage plans are a for-profit scam

@parismarx I think anyone who thinks AI will ever be used for good has never worked retail and had their schedule dictated by a computer system forcing them to clopen regularly but still giving them not quite enough hours to qualify for insurance.

@parismarx “our policy is for human oversight on AI decisions. Our training clearly states that it is the reviewer’s responsibility to validate the #ai findings. We have taken corrective action against the delinquent employees and added a new training program.” -company responding to these criticisms that expects same employees to review 1000s of cases a day.

@parismarx what the fucking fuck

@parismarx 90% error rate? That doesn't make sense. Did they forget to add a "!" before the last boolean?

@parismarx Sounds like its working as designed... #denied

@parismarx

That's why people need free government public service
For basic human health care which covers everybody.

There is no need to deal with technology / #Ai details where the main structure is obsolete.
if there is budgeting problem government can hire me as a consultant i will show them where to get it 🤐😉
Private insurance should be for jewelry etc. 😀
#HealhtCare

@parismarx Now look at the "error rate" for congress members need treatment. Do you think it's 90%? Not coincidence or happen stance they get better than us

@parismarx This doesn't even make sense. It's just bound to make the lawyers rich.

@parismarx 😮‍💨 Here we go again. This just sucks.

@parismarx wow. That is insane. Corporate greed plummeting health is terrible.

@parismarx Healthcare is a free market but people can't opt out of it or they will literally die.

@parismarx Strange problems that don't happen when there's Universal Health Care.

@parismarx Now we'll have to find new jobs for all those folks on the Death Panels...