Post

UHC is not in the healthcare business—-it IS in the claims denial business. AI , error rates or not, is just the latest tool in its arsenal.

seems like UHC still has some learning to do if one ceo wasn't enough.

this week's newsletter is all about this, ICYMI:

been telling reporter pals to watch for the head on collision between disintegrating rural healthcare and ai. the story will come twice. first as savior, then as failure.

Which is why it would be nice if the press could quickly steel itself to not using the industry's chosen nomenclature for a lot of this stuff, like "model misalignment." It's all very intentionally depersonalizing and distancing from responsibility

Confirmation bias in prompts plus misunderstanding, willful or not, the limitations of the system = profit. So what did UH and their competitors tell their claim review AI?

I see it in LLM coding use as well. The users seem to get a dopamine hit from immediate gratification, but also have an imaginary friend to pass blame onto. I’ve now heard “Claude must have…” excuses more times than I can count.

How would top diplomats, hostage negotiators, war crimes lawyers, primary school teachers, Nannies of the most rambunctious children, organized crime leaders, or regional community leaders negotiate with a rogue / misunderstood AI agent(s)? #Negotiation #Mediation #ArtificialIntelligence

Once again

AI does the one thing that really helps perpetrators of mass violence... okay two things it takes responsibility away from doing the horrible shit explicitly reinforces the frame of reference where killing for personal gain is being artificially lifted up as heroic, but the usa is expert in that

Cognitive surrender = moral surrender.

I have it on good authority UHC will employ heavy AI during the upcoming Annual Enrollment period for Medicare beginning 10/15. I would steer clear if possible.

I think it’s definitely this. They’ve taken an adaptive model and trained it to be highly error prone so that when the inevitable lawsuit hits they can point at some low level programmer and say “they did it” - then they just turn it back to 75% “errors” and say they fixed the issue.

The real horror isn't a superintelligence deciding we're obsolete; it's a mediocre algorithm deciding we're "optimized" into a permanent state of digital beige. We aren't being hunted by Terminators, just curated into oblivion.

Well, NerdBros can't engage with a self-aggrandizing John Conner power fantasy against late stage capitalist bureaucracy...

UHC denied a new CPAP to replace my defunct 10 yr old CPAP, because they claimed no record I have sleep apnea or used CPAP, even though they covered the sleep study, diagnosis, and my first CPAP. I am now without CPAP therapy after 10 yrs of 100% compliance.

Is this…AGI?

Yes, I was thinking AI is perfect for the Health Insurance Industry because a computer will never get tired of jerking people around unlike humans.

Even before gen-AI I liked to say that the threat from AI isn’t the Terminator, it’s a shit model that denies your mortgage approval

didn't they literally blow a guy's shit smoove off for this?

Is that by design?

For liability insurers, there are often severe bad faith penalties for wrongly denying coverage. It forces us to be very deliberate before denying anything. If we must have private health insurance, I don't get why we don't apply the same standards.

Apocalypse is the absence of any huge social movement fighting for a national public healthcare system in the US. Don't blame the machine by the lack of political consciousness of your countryman.

I am 65, w spinal condition & pain severe enough to prevent me from walking at all on the bad days. Blackbox AI program, NarxCare, told my new doctor that I'd become an addict by next week if I was given proper opiod medication. The fact that I successfully used the meds for 12 yrs was irrelevant.

In reality Kurt Vonnegut predicted all of this

Yup. The problem isn't the robots. The problem is who owns the robots.

How often Apocalypse arrives dressed just like Consequence.

That's probably not an LLM, but a simpler model of classification. And that has been a problem for a long time, so there are solutions available to stop it, like audits, legislation about bias, etc.

This article is from 2023. It’s old news.

There are videos of cops arresting wrong people for casino trespassing because their casino AI face recognition recognized wrongly, police trusted them

There's should be a 20% Interest rate that applies to all wrongfully denied claims (that starts accurring when the claim is filed and continues until the claim is paid). Also need some kind of independent public adjuster for health care claims.

Our Allina doctors were on strike this week with this as one of their pillars - keeping AI out of diagnoses. Like we want a machine involved in our health care??

bsky.app/profile/mada...

100% (even if things are very different now, 3 years later). If we set the bar at extinction or apocalypse, that lets the AI bros off the hook for all the dystopia their bots are still going to cause. "At least it didn't kill us all" is not nearly a good enough regulatory standard.

I mean, whether you're pushing the button to launch nukes or pushing the button to deny coverage...maybe it shouldn't be possible for any button-pushers to avoid accountability by saying "the button did it"

I know it’s a James Bond game but I think First Light is very prescient in having the primary A.I. danger being the lengths are willing to go to avoid admitting it could make mistakes.

But also... AI doomsayers aren't really a bunch I want discussing anything about healthcare. bsky.app/profile/segy...

Yeah, SkyNet is going to nuke the world because we put it in charge of the nukes even though its not capable of ruining a vending machine

Those aren't errors. The AI is doing what it is supposed to do: 1. Deny care as much as possible to maximize profits. 2. Take the blame as much as possible to maximize profits. That's it. A Magic 8-Ball could do it if the C-Suite thought it was really magic.

I have UHC they work. They suck but I don’t have a choice.

True, but also the venture capital private military startup industry is pretty terrifying and bleak. www.vcinfodocs.com

Yeah. All talk about apocalypse and nothing about the real harms happening now.

Of all the companies to do this....😬

Allegedly

UHC creating a computer program that does exactly what it wants is very bad, but it isn't a sign of an AI doomsday. It's just another sign that UHC is awful.

I don't respect your point.

You said, missing their point...

Insurance was never about protecting ordinary people. It's legalized extortion.

To their credit, they had a practice of denying care with a 90% error rate before AI, too (ask me how I know).

Luigi 2.0 when?

Bingo; people are expecting computer software to kill us via built machines because Movies Said So. The reality is, computer software will never kill us. But other humans using that software will, because they're either too corrupt or stupid to say "this software insisted I do something bad."

How could Luigi have done this

And they will likely argue that the person who gets fired if they disagree with the AI "too often" actually "makes" the decision to deny care. Our legal system needs to shut this kind of responsibility washing down, and hard.

Well, in this case these guys probably replaced humans who denied claims with 90% error rate with AI that does the same. It saved a bunch of humans from possibly feeling bad about being horrible people due to their job.

Before AI exterminates the world, it’ll just make things increasingly shitty for everyone first.

The only thing potentially worse than AI malfunctioning is that it works as intended

I am getting shit from the mods on the technology subreddit because I shared this article there despite it being three year olds and apparently as a result it went viral on X

Exactly. The media thickos want to focus on it destroying humanity because it’s Hollywood fiction which fuels clicks. In reality there are real harms right now. AI is shit. The fact that two software companies are effectively propping up the whole US stock market is the really scary part

As some who works at a dental office in the insurance/claims department, I’m not surprised whatsoever. This just confirms anytime they deny some of them more obvious important treatment

It’s not an “error rate” when that’s how they want it to work. It is, however, still negligence.

We know who to call

I bet Rick Scott likes this

The AI firms know this, happily taking advantage of the general public's ignorance of how computers work, much less LLMs.

kind of amazing to think this article got written without mentioning the POINT of AI was to fuck the system up and not pay out... it's not an "error rate"... it's a "non-indictable fraudulent process"

Uh, Yah. United Healthcare was a leader in denying care long before they used AI and now they have a higher rejection rate. @brooklynmarie.bsky.social

I would guess that UH would consider this 90% accurate. Anything in their favor is accurate.

When the error rate is that high, you may as well ditch the AI and pick randomly.

I think that's why the AI execs are pushing the narrative that the "AI escaped" so they can avoid responsibility.

As far as UHC is concerned, a 90% denial rate is a 10% error rate.

Luigi is always justified

Where's Luigi Mangione when you need him?

This story is almost 2 years old, if not older, and this tool is still being utilized

In conclusion - it would be better if we put our efforts into addressing the problem rather than screaming at each-other about which part of the problem is actually real and which is unimportant / sci-fi speculation /...