Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Improbable Lobster
Jan 6, 2012

"From each according to his ability" said Ares. It sounded like a quotation.
Buglord

Shame Boy posted:

at least i can sue a doctor for malpractice

There is some amount of accountability with a person, as opposed to the absolutely no accountability of software

Adbot
ADBOT LOVES YOU

Shaggar
Apr 26, 2006

mediaphage posted:

one interesting avenue of research on ai in medicine is some degree of guidance for doctors on treatment plans. not like where you let the ai doctor diagnose you and completely design your path through the healthcare system, but there are a lot of cases where somebody might be suffering multiple morbidities. and many times the “optimal” treatment plan for one condition will not only be contraindicated for another but also change how a third, fourth, or fifth might react. that’s a lot of information to keep straight in the moment

this doesnt need AI and portions/foundations of it already exist. the big problem is getting high quality, properly coded data into the system in the first place instead of the doctor just shoving everything into a notes field.

Shaggar
Apr 26, 2006

infernal machines posted:

okay, sure, but it must be worse than your average doctor somehow

lol

theflyingexecutive
Apr 22, 2007

Improbable Lobster posted:

There is some amount of accountability with a person, as opposed to the absolutely no accountability of software

every investor, board member, and c-level should get the ibm mantra tattooed on their arms for them to remember and on their foreheads for others to remember

Eeyo
Aug 29, 2004

i think one of my other problems is that llms and such aren’t “we can do it better” they’re “we can do it cheaper”. and if you follow that it has the inevitable conclusion of worse quality.

of course these days we just use money to determine who gets worse treatment. but you can squeeze even more people if you can make your ai reduce resource uses.

BMan
Oct 31, 2015

KNIIIIIIFE
EEEEEYYYYE
ATTAAAACK


Eeyo posted:

i think one of my other problems is that llms and such aren’t “we can do it better” they’re “we can do it cheaper”.

no, they just can't do it

Internet Janitor
May 17, 2008

"That isn't the appropriate trash receptacle."
"we can fake enough of it for a demo, faster and cheaper than any other prototyping method"

Improbable Lobster
Jan 6, 2012

"From each according to his ability" said Ares. It sounded like a quotation.
Buglord
i haven't really seen a path for chapgpt/llm ai to be both profitable and effective at whatever it's doing

Jonny 290
May 5, 2005



[ASK] me about OS/2 Warp

Improbable Lobster posted:

i haven't really seen a path for chapgpt/llm ai to be both profitable and effective at whatever it's doing

Firing a bunch of support people for a 1-3 quarter reduction in CoSt CeNteR budget to secure your bonus bag, after which you quit when the company falls apart. there you go.

Internet Janitor
May 17, 2008

"That isn't the appropriate trash receptacle."

Improbable Lobster posted:

i haven't really seen a path for chapgpt/llm ai to be both profitable and effective at whatever it's doing

especially right now when you're dependent on basically one irreplaceable third party vendor who can rugpull you at any time, and who you are by definition feeding training data and your "secret sauce" prompts to every time you use its apis

the security/cost calculus will change a bit when it becomes feasible to self-host everything, but llms are still going to be inherently unreliable pieces of poo poo

Zamujasa
Oct 27, 2010



Bread Liar

Jonny 290 posted:

Firing a bunch of support people for a 1-3 quarter reduction in CoSt CeNteR budget to secure your bonus bag, after which you quit when the company falls apart. there you go.

slash and burn departments with the promise of a new buzzword that will solve all the problems

Trimson Grondag 3
Jul 1, 2007

Clapping Larry

Improbable Lobster posted:

i haven't really seen a path for chapgpt/llm ai to be both profitable and effective at whatever it's doing

'compute will get cheaper over time' is not the worst risk you can take historically speaking but we are still at the point where compute power gains are getting taken up in pursuit of effectiveness.

defmacro
Sep 27, 2005
cacio e ping pong

NoneMoreNegative posted:



:eyepop: I wish you could read the whole thing without a signup bug in the way

jfc worst thing I've read in a minute. drat

Manzoon
Oct 12, 2005

ALPHASTRIKE!!!

Shaggar posted:

this doesnt need AI and portions/foundations of it already exist. the big problem is getting high quality, properly coded data into the system in the first place instead of the doctor just shoving everything into a notes field.

I helped my uncle one summer work on his fancy enterprise level dispatch software he spent way too much money on for his less than 10 person HVAC company.

He had someone a year or so prior to me showing up write out some very detailed steps on how to use it, inputting customer name, address, work done, etc. So instead of following that, they just put everything into the notes field. So I spent 8 hours a day for that summer just doing mindless data entry fixing their stupid database.

Powerful Two-Hander
Mar 10, 2004

Mods please change my name to "Tooter Skeleton" TIA.


Manzoon posted:

I helped my uncle one summer work on his fancy enterprise level dispatch software he spent way too much money on for his less than 10 person HVAC company.

He had someone a year or so prior to me showing up write out some very detailed steps on how to use it, inputting customer name, address, work done, etc. So instead of following that, they just put everything into the notes field. So I spent 8 hours a day for that summer just doing mindless data entry fixing their stupid database.

posted it before but oldjob spent a fortune on the Salesforce field level border encryption for sensitive fields and the sales guys just put all the sensitive info in the notes section that was unencrypted

Shame Boy
Mar 2, 2010

it says notes, i am going to write all my notes there, which is everything i would have otherwise written on a note pad, duh

mediaphage
Mar 22, 2007

Excuse me, pardon me, sheer perfection coming through

Shaggar posted:

this doesnt need AI and portions/foundations of it already exist. the big problem is getting high quality, properly coded data into the system in the first place instead of the doctor just shoving everything into a notes field.

using ml is fine and good in assisting medical decision making. also it’s literally impossible for doctors to easily make those decisions in a reasonable span of time. “the foundations” ah well that means it doesn’t actually exist.

anyway the problem of high quality data is part of it but that’s easier in canada

Trimson Grondag 3
Jul 1, 2007

Clapping Larry

Shaggar posted:

one of the biggest problems with EHRs is that they're designed to be customized for every provider that uses them which will always create bad outcomes, not just in terms of software quality but also in the poor medical processes they promote.

that said if traceable deaths to EHRs is only in the hundreds its probably doing waaaaay better than written paperwork.

last page but I just saw this. there was a huge privacy kerfuffle when australia implemented an opt out federally managed EHR system that all the care providers were supposed to input into so that health records would follow you etc. I think people were imagining a big sophisticated cross referenceable database but in reality it was a psuedo drop box wth 10dpi scans of your doctors illegible notes from 1994 with a bit of metadata. so of course no one uses it and they are going to relaunch the whole thing and try again.

kitten emergency
Jan 13, 2008

get meow this wack-ass crystal prison

Powerful Two-Hander posted:

posted it before but oldjob spent a fortune on the Salesforce field level border encryption for sensitive fields and the sales guys just put all the sensitive info in the notes section that was unencrypted

lol

Shaggar
Apr 26, 2006

Manzoon posted:

I helped my uncle one summer work on his fancy enterprise level dispatch software he spent way too much money on for his less than 10 person HVAC company.

He had someone a year or so prior to me showing up write out some very detailed steps on how to use it, inputting customer name, address, work done, etc. So instead of following that, they just put everything into the notes field. So I spent 8 hours a day for that summer just doing mindless data entry fixing their stupid database.

guess how 10000 employee provider orgs who spend a billion dollars a year on their EHRs handle the same problem? you'll be amazed!

Chris Knight
Jun 5, 2002

me @ ur posts


Fun Shoe

theflyingexecutive posted:

every investor, board member, and c-level should get the ibm mantra tattooed on their arms for them to remember and on their foreheads for others to remember
"no one ever got fired for buying IBM"?

"you can make lots of money selling systems to Nazis"?

help me out bro

Shaggar
Apr 26, 2006

mediaphage posted:

using ml is fine and good in assisting medical decision making. also it’s literally impossible for doctors to easily make those decisions in a reasonable span of time. “the foundations” ah well that means it doesn’t actually exist.

anyway the problem of high quality data is part of it but that’s easier in canada

why would it be easier in canada?

also ML for core medical decision making is always going to be sketchy both because of the black box nature and the total non existence of quality training data. its mostly a way to avoid having to document SOP in the first place. Instead of science based rules you write down in detail, you just create a black box trained on the garbage in existing EHRs. Its not totally worthless, but I would never in a million years trust that it doesnt miss loads of problems

infernal machines
Oct 11, 2012

we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

Chris Knight posted:

"no one ever got fired for buying IBM"?

"you can make lots of money selling systems to Nazis"?

help me out bro

anything involving mandatory arm tattoos and ibm is going to a bad place

Pollyanna
Mar 5, 2005

Milk's on them.


you’re thinking of ebm

Shaggar
Apr 26, 2006

Powerful Two-Hander posted:

posted it before but oldjob spent a fortune on the Salesforce field level border encryption for sensitive fields and the sales guys just put all the sensitive info in the notes section that was unencrypted

every time.

ADINSX
Sep 9, 2003

Wanna run with my crew huh? Rule cyberspace and crunch numbers like I do?

Chris Knight posted:

"no one ever got fired for buying IBM"?

"you can make lots of money selling systems to Nazis"?

help me out bro

I’m assuming they mean this:

haveblue
Aug 15, 2005



Toilet Rascal

Powerful Two-Hander posted:

posted it before but oldjob spent a fortune on the Salesforce field level border encryption for sensitive fields and the sales guys just put all the sensitive info in the notes section that was unencrypted

why didn't they encrypt the notes field

or would that have cost a second fortune

Shaggar
Apr 26, 2006

Trimson Grondag 3 posted:

last page but I just saw this. there was a huge privacy kerfuffle when australia implemented an opt out federally managed EHR system that all the care providers were supposed to input into so that health records would follow you etc. I think people were imagining a big sophisticated cross referenceable database but in reality it was a psuedo drop box wth 10dpi scans of your doctors illegible notes from 1994 with a bit of metadata. so of course no one uses it and they are going to relaunch the whole thing and try again.

tbf scanning in the documents and attaching them to the relevant patient and provider is a pretty big first step and if the goal was to have everything in the future be fully coded thats not a bad strategy under cost constraints. It probably sucks to search through the old records, but it sucks less than searching thru the old paper records.

The problem is if they keep doing it for new records.

Shaggar
Apr 26, 2006

haveblue posted:

why didn't they encrypt the notes field

or would that have cost a second fortune

if you encrypt the notes field we wont be able to share the notes with our business critical vendors

Beeftweeter
Jun 28, 2005

OFFICIAL #1 GNOME FAN
lol just grifting harder now

https://www.reuters.com/technology/openais-sam-altman-launches-worldcoin-crypto-project-2023-07-24/

quote:

OpenAI's Sam Altman launches Worldcoin crypto project

July 24 (Reuters) - Worldcoin, a cryptocurrency project founded by OpenAI CEO Sam Altman, launched on Monday.

The project's core offering is its World ID, which the company describes as a "digital passport" to prove that its holder is a real human, not an AI bot. To get a World ID, a customer signs up to do an in-person iris scan using Worldcoin's 'orb', a silver ball approximately the size of a bowling ball. Once the orb's iris scan verifies the person is a real human, it creates a World ID.

The company behind Worldcoin is San Francisco and Berlin-based Tools for Humanity.

The project has 2 million users from its beta period, and with Monday's launch, Worldcoin is scaling up "orbing" operations to 35 cities in 20 countries. As an enticement, those who sign up in certain countries will receive Worldcoin's cryptocurrency token WLD.

WLD's price rose in early trading on Monday. On the world's largest exchange, Binance, it hit a peak of $5.29 and at 1000 GMT was at $2.49 from a starting price of $0.15, having seen $25.1 million of trading volume, according to Binance's website.

Blockchains can store the World IDs in a way that preserves privacy and can't be controlled or shut down by any single entity, co-founder Alex Blania told Reuters.

The project says World IDs will be necessary in the age of generative AI chatbots like ChatGPT, which produce remarkably humanlike language. World IDs could be used to tell the difference between real people and AI bots online.

Altman told Reuters Worldcoin also can help address how the economy will be reshaped by generative AI.

"People will be supercharged by AI, which will have massive economic implications,” he said.

mystes
May 31, 2006

I think that was announced a couple years ago but I'm amazed it hasn't already died. I guess they're just pushing it as relating to ai now rather than crytocurrencies or nfts or whatever though?

kitten emergency
Jan 13, 2008

get meow this wack-ass crystal prison
orbing operations

haveblue
Aug 15, 2005



Toilet Rascal
:orb:

qirex
Feb 15, 2001

the only salesforce instance I've ever seen where the users didn't just put data wherever they wanted was in wealth management where a privacy fuckup would be career ending

Shame Boy
Mar 2, 2010

qirex posted:

the only salesforce instance I've ever seen where the users didn't just put data wherever they wanted was in wealth management where a privacy fuckup would be career ending

i've seen it even in wealth management, back when i used to have to deal with that

Shame Boy
Mar 2, 2010

this was way back two companies ago so idk maybe it got better but i'm gonna guess not.

they would also forward us email chains sometimes that included people's entire files including SSN and account numbers and balances which was pretty lol

theflyingexecutive
Apr 22, 2007

ADINSX posted:

I’m assuming they mean this:



yes this one

qirex
Feb 15, 2001

Shame Boy posted:

this was way back two companies ago so idk maybe it got better but i'm gonna guess not.

they would also forward us email chains sometimes that included people's entire files including SSN and account numbers and balances which was pretty lol

this was a different kind of place, like you could be fired for just attempting to search for an important client's name even though you didn't have access to their profile. also the really big deal people didn't have their real names in the system, they had aliases

mediaphage
Mar 22, 2007

Excuse me, pardon me, sheer perfection coming through

Shaggar posted:

why would it be easier in canada?

larger quantities of high quality centralized medical records thanks to public healthcare. you still have to go through clean up but there’s a lot of very good data

and this isn’t about making something make all the decisions for you, it’s more about elevating causes you might miss because your patient has six illnesses. i’m speaking specifically about its use in easing diagnostic burdens in multimorbidity cases.

Adbot
ADBOT LOVES YOU

mediaphage
Mar 22, 2007

Excuse me, pardon me, sheer perfection coming through
lol how’s a new crypto coin going to do anything to tell the difference between humans and non-human content. what’s preventing a human from loaning its id or more likely faking it in the first place

also loving lol at giving a private crypto company my goddamned iris scan

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply