> Increasingly health insurance companies and healthcare providers are intertwined.
Kaiser is an HMO. They are the insurer and try to have their employees, such as these nursing lines, perform almost all of their care. Shifting from one line of business to another is purely internal and can't game Medical Loss Ratio like your scenario.
> they don’t have much incentive to lower total cost because 20% of a larger number means more total profit
There is some bad incentive here for sure. That being said, insurers do compete on price so they lose customers if they charge more than other insurers. Also, regulatory rate review can decide whether they can raise premiums a given amount.
The article doesn't seem to be about replacing nurses staffing their phones with AI. It seems to be about making sure the care they're providing is what the company wants.
I'm not sure the for-profit approach is exactly what's to blame. HMOs like Kaiser are legally forced to spend a certain fraction (80% percent for the worst case, more for large group plans) of their premium revenue on medical services. They can't save and pocket the money like a traditional for-profit enterprise.
This doesn't seem like a money-saving measure exactly. The main AIs the article talks about is making sure nurses on their nurses' lines aren't being assholes. I guess this used to be spot checked before so you save on that? Maybe? It seems like they are trying to solve the problem of some of their nurses staffing their nurses' line not treating their patients the way they're supposed to.
I'm getting a lot of refusals these days from multiple LLMs on multiple fronts for silly stuff, a lot more than I had for a while. If this is where things are really going, I think open weight models have a big future.
That I don't know if they're a stranger or not. I introduce myself to acquaintances fairly regularly (sometimes annoying them that I apparently think they're so unmemorable) because I'm the opposite of 'I never forget a face'.
I think normal people are more likely to have the experience where they can't remember a name and why/how/where they know someone from. I of course forget things like anyone, but that's unrelated.
Like the poster, I'm faceblind. It isn't the worst thing: I'm not voice blind, height blind, age blind, hairstyle blind, gender blind, features associated with race and ethnicity blind, attractiveness blind, affect blind, context blind, etc., so I'm mostly good at figuring out who someone is. Within one encounter with a bunch of people, I try to note what someone is wearing.
Every once in a while I don't recognize someone and I go through this whole thing of bringing up every biographical detail about them I remember and all the things we've talked about to show that I'm not an asshole who wasn't paying attention in the past. Fortunately, I have a decent memory for such things.
LLMs are great with minority languages compared to almost anything else. Including better than the by the natural language generation employed to use Abstract Wikipedia, which whiffs at relatively large languages like Zulu and Xhosa, let alone many of the rarer languages that popular LLMs speak fluently.
Other correctly point out it does matter what language the code is in since the human does sometimes need to read and understand it.
But also, I suspect the article is just wrong. "The hard languages got easy first" isn't true in practice and the impressive examples given are not representative or as magical as the poster makes them out to be.
The takeaway might be right in the end, but the post isn't right in the beginning.
> Physicians use all their senses. They poke, they prod, they manipulate, they look, listen, and smell.
Sometimes. Sometimes they practice by text or phone.
> They’re also good at extracting information in a way that (at least currently) sycophantic LLMs don’t replicate.
If I had to guess, I think I'd guess that mainstream LLM chatbots are better at getting honest and applicable medical histories than most doctors. People are less likely to lie/hide/prevaricate and get more time with the person.
Would be interested to hear a legal expert weigh in on what 'advice' is. I'm not clear that discussing medical and legal issues with you is necessarily providing advice.
One of the things I respected OpenAI for at the release of ChatGPT was not trying to prevent these topics. My employer at the time had a cutting-edge internal LLM chatbot for a which was post-trained to avoid them, something I think they were forced to be braver about in their public release because of the competitive landscape.
I'm struggling to understand what the result really is: it seems that some dogs at some point would rather play with a toy than eat or come play with their owner. That seems pretty normal. Is this really "addictive-like"? Why isn't it "really enjoy"?
Whenever I try to read up on it, it seems like glaciers are receding at ~2x their without-climate-change rate. That's a huge increase, but it doesn't seem like there's something that a person can experience at a visceral level here that is based on fact and not just preconception.
It's definitely striking, I can't deny that. I crossed the last remnants of an almost-extinct glacier last year that my guide guessed would be gone in 1-3 years: at the beginning of his career it was a real glacer with non-trivial extents, crevasses, etc.
Yes, there are unilateral policies and treaties that let the US and the UK collaborate in legal action (going through US institutions to judge them), some of them referenced in https://meta.wikimedia.org/wiki/Legal/Legal_Policies -- a keyword might be letters rogatory
Wikimedia also seems to have a presence in the UK https://wikimedia.org.uk/ that presumably would be affected.
In most cases they might have enough pull to get folks blacklisted by payment processors, but wikimedia in particular might win that one.
with sql_context(columns="x"):
query, values = sql(t"SELECT {col} FROM y")
I think
1. this is relying on the `col = "x"` in the previous example
2. columns is a set of strings, so it might be sql_context(columns={"foo", "bar", "x"}) to allow those as valid options. It just happens that "x" is a collection supporting the `in` operator so it works much like the set {"x"} would.
2a. (You might hope that something would convert such a string to a singleton set, but I don't think it does, which would have weird results with a multi-letter string.)