(Thread IKs:
skooma512)
|
Paradoxish posted:Don't ban cars going into philly because then my business trip at the end of the month will literally cost me an extra $300. In the Seattle area we have a commuter heavy rail train called the Sounder but because the tracks it runs on is owned by BNSF, who charges a king's ransom for usage, it only runs for morning and evening commute times M-F, so I can't use it to get anywhere on the weekend or evenings.
|
#
?
Jun 14, 2023 06:11
|
|
|
|
| # ? Jun 11, 2024 18:04 |
|
|
The patient facing "AI" in health care is going to be a chat bot that is integrated into the patient portal that can answer basic questions about office locations and copays and all of that. It may be mildly helpful. The "AI" that everyone wants is something that you can feed the entire patient chart into and it will do... something?? Everyone is slobbering to do this, despite the fact that we've spent the last twenty years and billions of dollars to make the chart discrete, let's just ignore that and cram our progress notes into the mystery box and see what happens.
|
|
#
?
Jun 14, 2023 06:20
|
|
|
Lacrosse posted:In the Seattle area we have a commuter heavy rail train called the Sounder but because the tracks it runs on is owned by BNSF, who charges a king's ransom for usage, it only runs for morning and evening commute times M-F, so I can't use it to get anywhere on the weekend or evenings. There's a solution for it. like Capitol corridor you contract it to amtrak and then the his railroad can't say poo poo. but sound Transit is more interested in stalling projects for decades
|
|
#
?
Jun 14, 2023 06:54
|
|
|
They spent 30 years adding an extra few lanes between JBLM and Tacoma. Usable rail is near impossible and will require at least a decade of feasibility studies
|
|
#
?
Jun 14, 2023 06:56
|
|
|
BULBASAUR posted:They spent 30 years adding an extra few lanes between JBLM and Tacoma. Usable rail is near impossible and will require at least a decade of feasibility studies "we can't have rail because that will be chinese authoritarianism" as i democractically grift billions
|
|
#
?
Jun 14, 2023 07:17
|
|
|
BULBASAUR posted:They spent 30 years adding an extra few lanes between JBLM and Tacoma. Usable rail is near impossible and will require at least a decade of feasibility studies It did nothing for traffic, still is a massive choke point
|
|
#
?
Jun 14, 2023 07:23
|
|
|
nikosoft posted:The patient facing "AI" in health care is going to be a chat bot that is integrated into the patient portal that can answer basic questions about office locations and copays and all of that. It may be mildly helpful. pray the mystery box never gives you a misdiagnosis because that will stick with you forever too
|
|
#
?
Jun 14, 2023 07:51
|
|
|
Glumwheels posted:It did nothing for traffic, still is a massive choke point Its honestly a disgrace. The dilapidation in this country is astounding
|
|
#
?
Jun 14, 2023 08:01
|
|
|
Jel Shaker posted:pray the mystery box never gives you a misdiagnosis because that will stick with you forever too My body is a mystery box anyway so what's another one
|
|
#
?
Jun 14, 2023 08:53
|
|
|
FizFashizzle posted:if AI ever does make serious headway into the medical field it will be for poor people without insurance. It will join the likes of chiropractic medicine and aromatherapy; ineffective alternatives for people that are getting hosed over by the system. This is 100% the call and honestly for a lot of people it might be better than what they have now? Is that a nutty opinion? I have known people in my life who are scared to talk to a doctor. I have known people who find it "too real" to go see a doctor for a big problem and would rather let it get really bad and force the issue. I have known people who can only afford the kind of care where you feel like cattle and are ushered through a lovely diagnostic where no one takes you seriously, especially when it would require more money. These models will suck, but would they be worse than a random person googling and then making up their own medical diagnoses? Cuz to me, that's the alternative. Just having some nominally authoritative way to receive a quick, free "diagnosis" feels like it will be better than nothing for people who are, for whatever reason, not enmeshed in the system or otherwise thinking about their long term goals. Which has always been one of the worst problems of poverty right? You can pretty much guess the financial ability of a human being by how much they intentionally arrange and plan for the future. Not just because poorer individuals don't have the time for introspection, but because they're rightfully scared of thinking about their future in a world that left them behind. Maybe under those conditions having a bot who can provide some kind of "better than nothing" diagnosis is preferable. The stratification of society continues and once again Idiocracy gets it right. If anything I'm struck by your cheery outlook that such a thing only "might happen". To me it feels inevitable, and like everything else, it will start at the bottom 25% and insidiously trap the bottom 75%+ over time. Taima has issued a correction as of 09:32 on Jun 14, 2023 |
|
#
?
Jun 14, 2023 09:21
|
|
|
isn't an AI's inability to "be wrong" a huge problem if you're using it for something as serious as a medical diagnosis? like, if you feed it a bunch of symptoms and a patient history, and ask it for a response, and it can't find anything that directly points to the answer, it's going to make something up, in much the same way that the dude who tried to use chatGPT to make a legal argument ended up citing cases that literally did not exist and isn't that going to just straight-up kill people?
|
|
#
?
Jun 14, 2023 09:38
|
|
|
sure, but they scrolled through and accepted a 100 page EULA before speaking with the bot
|
|
#
?
Jun 14, 2023 09:41
|
|
|
you just set the confidence interval coefficient of MEDgpt5.3 to 0.6x instead of 1.0x I'm mostly joking but I imagine the hope is they will be wrong less frequently than the arithmetically mean human doctor  e: I have also started seeing a 40-something nurse practitioner (with a masters I think?) and it's been decent. not good but definitely clearing the low bar of "best care" in this country. The hospital/clinic system is likely to be or already has been purchased by some spurious religious entity though so I may have to change providers. My employer is slow to respond since they only reevaluate insurance annually. cool, love the system. I'd like to order 3 meatballs, and then 3 meatballs, and then clear the plates and my wife will show up and I'll tell you my order. It's meatballs. palindrome has issued a correction as of 09:50 on Jun 14, 2023 |
|
#
?
Jun 14, 2023 09:42
|
|
|
Yeah, in a weird way being wrong matters less with medical care, because the current system already gets it wrong on a regular basis. A replacement just has to be less wrong, gently caress up fewer people's lives, to be an acceptable alternative. And if you're talking about uptake instead of societal impact, look at how many people flock to alternatives that are always wrong, like "actually, your children should drink
|
|
#
?
Jun 14, 2023 09:55
|
|
|
Good points, and let us not forget the potential cost savings for tele-health appointments with a bot instead of a person on a conference bridge where someone struggles to get their browser to join and audio working for the first few minutes
|
|
#
?
Jun 14, 2023 10:03
|
|
|
Yeah, I don't think a bot spitting out answers is that great especially since it could easily go down the wrong direction as far as symptoms go. You want at least someone doing a little bit of critical thinking, not that it isn't going to happen but I could see people getting actively detrimental care out of it.
|
|
#
?
Jun 14, 2023 10:59
|
|
|
"The system is broken. Inequities persist in our egalitarian wasteland. We must feed all those codified inequities into a heap and blessed AI will chat our booboos away."
|
|
#
?
Jun 14, 2023 11:00
|
|
|
Mr Hootington posted:A person with a Yeti cooler is signaling their genetic material could produce a high IQ child They store their spare genetic material in the Yeti cooler in case its ever needed.
|
|
#
?
Jun 14, 2023 11:11
|
|
|
https://twitter.com/WSJ/status/1668864610978484225?s=20
|
|
#
?
Jun 14, 2023 11:21
|
|
|
(intense mariachi guitar music)
|
|
#
?
Jun 14, 2023 11:40
|
|
|
Modelo is also owned by Inbev, it is literally one division overtaking another division.
|
|
#
?
Jun 14, 2023 11:42
|
|
|
Ardennes posted:Modelo is also owned by Inbev, it is literally one division overtaking another division. The spice must flow...
|
|
#
?
Jun 14, 2023 11:55
|
|
|
I’ll still keep drinking that garbage
|
|
#
?
Jun 14, 2023 12:16
|
|
|
Ardennes posted:Modelo is also owned by Inbev, it is literally one division overtaking another division. lol
|
|
#
?
Jun 14, 2023 12:17
|
|
|
gradenko_2000 posted:isn't an AI's inability to "be wrong" a huge problem if you're using it for something as serious as a medical diagnosis? https://twitter.com/MIT_CSAIL/status/1604884273789603842 also the lawyer thing is even stupider, the one guy asked it to generate some case law, so it did, and he clearly didn't read it at all because it was completely incoherent between paragraphs. then he "verified" it by asking chatgpt if this was a real case, and it said yes, so he did the same with a bunch more. then his boss just blindly signed off on the whole thing. if they put this thing in any part of the prescription chain it will be really bad
|
|
#
?
Jun 14, 2023 12:19
|
|
|
I've watched doctors -- specialists even -- Google poo poo right in front of me so I don't fear the AI in that sense. I don't see how it would do much for inequality of care though. The process for trying to escalate something beyond the AI to get an actual prescription or procedure will probably be hell.
|
|
#
?
Jun 14, 2023 12:19
|
|
|
Chad Sexington posted:I've watched doctors -- specialists even -- Google poo poo right in front of me so I don't fear the AI in that sense. orthopedics doc pulling up a big animated gif of rotator cuff would have been way more helpful than him grabbing a printout of shoulder anatomy and drawing on it, but they don't have screens in the patient rooms. e: i cut my leg and it wasn't getting any better with neosporin and bandages, was constantly red and itchy. went to doc asked if it looked infected, she asked what i was doing and I said neosporin and she immediately goes "i hate neosporin here's a prescription for something else" and it got better real fast. apparently I spontaneously developed an allergy to the thing in neosporin that 10% of the general public is allergic to. AI would have just said "apply neosporin and bandages" Griz has issued a correction as of 12:29 on Jun 14, 2023 |
|
#
?
Jun 14, 2023 12:25
|
|
|
Chad Sexington posted:I've watched doctors -- specialists even -- Google poo poo right in front of me so I don't fear the AI in that sense. a doctor googling in front of you is a million times better than a doctor just making stuff up and a thousand times better than the average person because they have a better idea of what to look for than just to go straight to cosmopolitan health advice section
|
|
#
?
Jun 14, 2023 12:25
|
|
|
It's Modelo time foo'
|
|
#
?
Jun 14, 2023 12:27
|
|
|
 
|
|
#
?
Jun 14, 2023 12:30
|
|
|
turns out it was a masterful tactic to shift production of beer to Mexico all along
|
|
#
?
Jun 14, 2023 12:33
|
|
|
Jel Shaker posted:a doctor googling in front of you is a million times better than a doctor just making stuff up and a thousand times better than the average person because they have a better idea of what to look for than just to go straight to cosmopolitan health advice section more like "doctor looking up poo poo on the hospital system intranet" since the internet in general is about to be poisoned with infinite amounts of AI garbage from the content farms that already mastered SEO optimization
|
|
#
?
Jun 14, 2023 12:33
|
|
|
it’s amazing watching everyone destroy the internet
|
|
#
?
Jun 14, 2023 12:36
|
|
|
Griz posted:more like "doctor looking up poo poo on the hospital system intranet" since the internet in general is about to be poisoned with infinite amounts of AI garbage from the content farms that already mastered SEO optimization true but most of the time it’s just using the google search engine to look at a few well known websites such as Uptodate, GPnotebook or if you’re feeling fruity Radiopedia pubmed is also a great resource but half the time the article you want costs $$$ so you just open your phone and hope the scihub mirror isn’t blocked by IT yet
|
|
#
?
Jun 14, 2023 12:39
|
|
|
euphronius posted:it’s amazing watching everyone destroy the internet message boards somehow have a future, hell, there are absent goons that have returned because reddit's killing itself
|
|
#
?
Jun 14, 2023 12:41
|
|
|
Griz posted:https://twitter.com/MIT_CSAIL/status/1604884273789603842 The lawyer sure was stupid, BUT, I'm actually kinda sympathetic to him. Apparently, the same corporate entity that gates public funding science via exorbitant journal licensing fees ALSO owns & operates the case file access system (with opinions and rulings produced in public courts by publicly funded judges). Law practitioners need this, but unsurprisingly it similarly requires expensive licensing fees to access. GPTIdiot lawyer's firm usually only did state law cases and therefore ONLY PAID FOR STATE ACCESS. He was tasked with finding case law to support a federal case 1) he couldn't argue himself because he wasn't licensed because unfamiliar with the procedures, and 2) in a firm that refused to pay for the proper tools to do so (partly due to highway robbery on the part of the "publishers"). Simultaneously to that, there's been a legitimately huge push by people who really should know better to spread wildly inflated impressions of what this tech will do, with minimal warning language that is 1) not super clear especially to non-technical people, and 2) as far as I can tell changes without notice over time, I have no idea what warnings actually were present when he signed up for OpenAI. He did notice incoherence between paragraphs and attributed it to ChatGPT returning "excerpts" (a conclusion I have no doubt chatGPT would have concurred with and probably did, as he likely mentioned this to the tool he thought was a legit AI, in line with what he was TOLD it was.) His boss signing off on it blindly was pretty stupid to be sure. But then again remember this was all happening in a firm that was not providing proper tools for the work. Frankly, dude is extremely lucky he is in a profession where his work actually is checked to such a degree and his misperceptions identified and corrected relatively early on in the process. The slow horror is the certainty that some professional mechanical engineer somewhere has used it to look up the tensile strength of X, Y, & Z materials he's unfamiliar with due to similar lack of easy access to the normal databases or tools, as it's kinda related to his normal work but not really an area he's got much real experience in. The resulting bridge/electrical system/water filtration pumps are probably just now gearing up for construction so let's hope those errors get found & corrected before it kills a bunch of people. But law being inherently adversarial with humans on both sides means that errors are MOST likely to be caught early in that field. Think we're eventually going to get a nice long, slow, but unfortunately devastating, lesson on what these "AI tools" do in other fields. jemand has issued a correction as of 12:54 on Jun 14, 2023 |
|
#
?
Jun 14, 2023 12:50
|
|
|
zetamind2000 posted:message boards somehow have a future, hell, there are absent goons that have returned because reddit's killing itself I mean, we're all dying all the time.
|
|
#
?
Jun 14, 2023 12:55
|
|
|
zetamind2000 posted:message boards somehow have a future, hell, there are absent goons that have returned because reddit's killing itself Anyone who tries to crawl back from Reddit should be permad
|
|
#
?
Jun 14, 2023 13:03
|
|
|
AI in legal searches is completely useless.
|
|
#
?
Jun 14, 2023 13:16
|
|
|
|
| # ? Jun 11, 2024 18:04 |
|
|
jemand posted:Frankly, dude is extremely lucky he is in a profession where his work actually is checked to such a degree and his misperceptions identified and corrected relatively early on in the process. The slow horror is the certainty that some professional mechanical engineer somewhere has used it to look up the tensile strength of X, Y, & Z materials he's unfamiliar with due to similar lack of easy access to the normal databases or tools, as it's kinda related to his normal work but not really an area he's got much real experience in. The resulting bridge/electrical system/water filtration pumps are probably just now gearing up for construction so let's hope those errors get found & corrected before it kills a bunch of people. But law being inherently adversarial with humans on both sides means that errors are MOST likely to be caught early in that field. Think we're eventually going to get a nice long, slow, but unfortunately devastating, lesson on what these "AI tools" do in other fields. just remembered that problem with a photocopier replacing all the values on architectural plans with the same number because it's a digital process now where the machine takes a picture of the document and then reprints it, but the compression algorithm decided that everything looks similar enough to 24.01 that you might as well make all of them read that
|
|
#
?
Jun 14, 2023 13:31
|
|