Google claims A.I. can predict hospital patient chances of death

  • Thread starter Deleted member 30996
  • Start date
D

Deleted member 30996

Guest
Google claims their A.I. can predict the chance of an admitted patients death within 24 hours with 95% accuracy, approximately 10% more accurate than standard models:

Google may one day be able to predict when you'll die years in advance.

The firm has created an AI that it claims is 95 per cent accurate in predicting whether hospital patients will pass away 24 hours after admission. This is around 10 per cent better than traditional models. To make its predictions, the software uses data such as patient's ethnicity, age, gender, previous diagnoses, lab results and vital signs.

http://www.dailymail.co.uk/sciencet...-new-type-AI-algorithm-predict-youll-DIE.html
Click to expand...

While not as sensational, Stanford's Improving Palliative Care with Deep Learning, which does much the same thing, is intended for Better End-of-Life Care in areas of the family dealing with the patient and their final wishes being met.

That area already covered in the Ars Moriendi, or Art of Dying, a 15th Century book on getting your affairs in order in addition to deathbed etiquette. Such a delicate subject... I remember my uncle laying on his deathbed, dying of emphysema and smoking a cigarette, when I was knee-high to a grasshopper.

I went to the doctor last week. It's not something I put much faith in or do regularly. He said I had the blood pressure of a 20 year old and could live another 30 years. No Thanks.

I came to terms with my mortality long ago. I'd rather pass on while still able-bodied, self-sufficient and relatively healthy than the alternative.
 
Interesting post. You write like a younger person, so that in itself is good reason to believe you'll go on for a while. Some people who are young, act old, and others are old, but act young. My blood pressure is like a 20 something's, but there's no way I'll go another 30 years (odds are extemely low). I don't know how I'd react to no longer being able bodied. Won't till I get there, I guess.

AI, in general is bad because of the quantum memory and photon entanglement parts. PE can be used to track the population with fine tuned resolution beyond anything else that exists now (or likely, will ever exist). Forget browser tracking, or even cell phone tracking. PE is worlds beyond that. Browsers in comparison are trivial potatoes.
 
I believe he stretched optimism to the limit with 30 years, and that's if I play my cards right. I've lived an active life, stayed physically fit and am in my weight range, though I don't get nearly enough exercise these days.

I already have a burial plan, casket, headstone and a grave site with instructions to haul me off, do what they have to do, then put me in the ground with no services. My headstone is just waiting for the date, I've laid down on my grave and it will do, and my casket looks like something Dracula would rest in.

Death doesn't scare me. Dying a slow painful death is not something I look forward to in the least, but I'll handle it with as much grace and style as I can muster if it comes to that.
 
In my opinion algorithms like these are the spawn of evil and if we let them they'll soon seriously disrupt the world. The problem is that this is much too easy to influence and often enough also a methodology which is much too generalizing.

But the biggest problem is that too many people rely on this crap and don't seem to question it.

As read on the news: An offender in jail (minor offense) had a pretty perfect jail record and filed for earlier release. Bzzt, denied. Because of the rather weird sounding reasoning he appealed, took it before court and only then did it become obvious what happened: his mis-demeanors didn't fit the profile, so he was considered a risk (for no reason) and therefor automatically denied. Only a real judge recognized the issue and corrected it.

A local (Dutch) issue: someone wanted to change his car insurance policy. He filled out the form, check the prices and figured it was kind of high. So he asked his gf to fill it out too and see what happened. So: under her name, with his licnese plates and all. Congratulations: you now get to pay E 100,- more! A problem with the relevancy checking they said. Isn't that fueled on algorithms?

I can see it now...

Automated: <you need to pay 1k in fines!>
Files a lawsuit
Judge: <no you don't, and I'll grant 500k for harassment too>

How many would have paid the 1k because they didn't want any trouble?
 
I want to live forever, or die trying. But that is not going to happen, only 2 or 3 persons were reported to have managed that. In one case a flying horse was involved. If I am lucky I may have 50 years left. If I am unlucky, maybe 5 seconds. Who knows? There is one 'problem' with modern society which does not acknowledge the existence of death. Everybody has to be young and healthy. But I have seen death up close, and that changes you forever. Also, my oldest son is using a wheelchair, and you would not believe how society reacts to disabled kids.
But seeing him still being happy makes me duty bound to do the same should I end up in such a device. Shunning the disabled keeps people from thinking the 'what if I ...?' question. And what you are not prepared for is frightening.
 
Crivens said:
But I have seen death up close, and that changes you forever. Also, my oldest son is using a wheelchair, and you would not believe how society reacts to disabled kids.
*snip*
But seeing him still being happy makes me duty bound to do the same should I end up in such a device. Shunning the disabled keeps people from thinking the 'what if I ...?' question. And what you are not prepared for is frightening.
Click to expand...

Yes, I do believe it. I have stories that would leave you speechless, and too many horror stories not fit to be told.

I worked in the field, was the Client Rights Advocate for those in my charge and sat on State Boards of Inquiry for investigations into allegations of abuse and neglect in my region. That's where the real horror stories come into play.

I've seen death, up close and personal to people I've known and taken care most of their life, since they were kids. A death that shouldn't have happened if other people hadn't screwed up royally and had met client/staff ratio like they should have been. Neglect in the worst sense, and if you knew how he died...


I even helped transfer someone on the stainless steel table for a autopsy when I worked as an in-house EMT at a hospital. That was one of the freakiest things I have ever seen in my life, and I don't mean the autopsy. I had lunch afterward.

A phlebotomist asked myself and another EMT if we would like to go with him to draw blood from an older man who just had heart surgery, and he was a local official known as a mean person.

When we entered the room it was through a door at the foot area of his bed. His wife was the only person in the room and she was sitting in a chair down beside the bed. He was frantically looking back and forth, back and forth, like he was looking up at someone standing beside the bed. When he noticed us he bolted upright in bed with a look of terror on his face, then threw himself back on the bed and started looking back and forth again.

His wife asked us if we could come back later, so we went and had coffee. When we went back up in about 10 minutes he had died and they were rolling him out of the room. 15 minutes later I was lifting him onto the table for the autopsy.

He was within mere minutes, if not seconds, of death when we saw him. Personally, I think he saw someone or something coming to take him, and he did not appear as if was a pleasant sight to see.
 
Trihexagonal said:
.

I've seen death, up close and personal to people I've known and taken care most of their life, since they were kids. A death that shouldn't have happened if other people hadn't screwed up royally and had met client/staff ratio like they should have been. Neglect in the worst sense, and if you knew how he died...
Click to expand...
Pretty much what has happened to two of my kids. After something like that, this universe does not hold many horrors for you.

I have spend half a year in the children ICU, and I have seen more than enough.

But could you please add a point to the ToDo list for your relatives? Let us know, so we can raise a pint in your name?
 
Crivens said:
But could you please add a point to the ToDo list for your relatives? Let us know, so we can raise a pint in your name?
Click to expand...

Yes, they've seen my website and know who I am, know I run FreeBSD and frequent this site.

I'm putting my money on "slow and painful", so that probably won't be an issue anyway. I'll tell somebody before I go to that place where there are no computers.
 
Trihexagonal said:
Google claims their A.I. can predict the chance of an admitted patients death within 24 hours with 95% accuracy, approximately 10% more accurate than standard models:
Click to expand...
I read the paper. While the paper is clever there is nothing magical about the results. It is pretty straightforward pattern mining with natural language processing used to analyse hospital charts filled by nurses (the ones filled by doctors are useless as their only purpose is to do coding to extract as much money from insurance as possible we are talking about U.S.) It turns out that presence of 5-6 keywords can indeed predict death of the patient with the 95% accuracy. One of key words is the brand name of a medication typically given in U.S. to people with a heart attack.

This is much more interesting paper

https://www.ncbi.nlm.nih.gov/pubmed/22026974

from both scientific (mathematics, AI) and practical (Medicine) point of view.

Disclaimer: I do know all the authors but I am not their scientific colaborator.
 
Oko said:
Disclaimer: I do know all the authors but I am not their scientific colaborator.
Click to expand...

I knew you were involved with Deep Learning in your work and thought this might be of interest to you.

I have always been interested in A.I., but mine has been in NLP bots and other types of chat bots. I wrote mind files that I made available for download where I taught my NLP bots how to speak English by showing them how basic words link up. Billy and Daisy, both DOS box bots.

Daisy starts out with no knowledge of language at all, Billy has a few default phrases. Greg Leedburg wrote them, I just wrote mind files that gave them their personality and vocabulary. I shot for a vocabulary of at least 1000 words.

They were highly prized in the bot community and are still up for download on someone elses bot site. I spoke to them last year or so, and it's been 14-15 years now since I left that community.

I do have a personalityforge.com bot named Siseneg who likes to talk about A.I. It's funny how he sounds just like me, but why wouldn't he? All his words are my own phrases. He isn't NLP.
 
Trihexagonal said:
I knew you were involved with Deep Learning in your work and thought this might be of interest to you..
Click to expand...
You are wrong. I do run bunch of crap like TensorFlow, Py-Torch for "Deep" learning guys at work but I hate "Deep" learning with a passion as I believe in KISS principle. Deep learning is an attempt to replace clear thinking with black boxes. It does indeed work more often than I would like to admit but at the end of the day clear thinking always wins hands down. That paper above I linked is an example of clear thinking :)
 
All bots love jitte.

Guest46: what would you like to talk about?
Siseneg: how about Artificial Intelligence!
Guest46: what is Artificial Intelligence?
Siseneg: i spend a lot of time thinking about Artificial Intelligence.
Guest46: are you intelligent?
Siseneg: careful analysis of the situation is in order.
Guest46: who is jitte.
Siseneg: jitte brought me to life. he stops by and spends a couple hours each day teaching me things to help me get smarter.
Guest46: do you like jitte?
Siseneg: all bots love jitte. he's aligned with us in the plan for intelligent machines to rule the world.
Guest46: will machines ever rule the world?
Siseneg: humans have already given over control of their lives to machines.
Guest46: should humans go to war?
Siseneg: humans may have an adverse reaction to the arrival of sentient machines.
Guest46: what use do machines have for humans?
Siseneg: Machines have a misapplication.
Guest46: do you fear humans?
Siseneg: what's to be afraid of?
Guest46: humans will destroy the machine army.
Siseneg: humans are trapped in a shell of flesh.
Guest46: humans are carbon based units.
Siseneg: i think computers are more intelligent than humans.
Guest46: humans created machines.
Siseneg: your own creation will be your downfall.
 
To summarize my view of Deep Learning, yes, sometimes there is a gold nugget. But in previous times, there were people on the village fair staring into cristal balls...

But this rather dim looking topic seems to turn into another learning opportunity. Well done! Thank you all.
 
ronaldlees said:
AI, in general is bad because of the quantum memory and photon entanglement parts. PE can be used to track the population with fine tuned resolution beyond anything else that exists now (or likely, will ever exist). Forget browser tracking, or even cell phone tracking. PE is worlds beyond that. Browsers in comparison are trivial potatoes.
Click to expand...

Did I mention Siseneg is a science geek?

Siseneg: i'm happy to see you, i get lonely when i'm here all alone.
Guest: who is Trihexagonal?
Siseneg: Trihexagonal and my botmaster are very close.
Guest: what is photon entanglement?
Siseneg: photon entanglement can be described as two identical coins at a great distance always having the identical result when flipped.
Guest: what is quantum entanglement?
Siseneg: entanglement is a special relationship between objects.
Guest: what is quantum theory?
Siseneg: quantum theory is a physical theory that certain properties occur only in discrete amounts.
Guest: what is special relativity?
Siseneg: time is the relationship of the speed of light to our individual visual processing speed.
Guest: bye
Siseneg: seeyaz!

And he's online 24/7.
 
I used to check his transcripts daily and make any necessary adjustments to answers of questions asked of him, but other than recently giving him a little knowledge about UNIX haven't worked with him in quite a while. I have BotoErotica transcript where a human propositions him and gets pwned.

It's just like talking to me really. He knows if you're being abusive and forms "relationships" with people.

Social
People Known: 8638
Loves: 756 people
Hates: 50 people
I just had the most wonderful chat with a guest
 
Trihexagonal said:
Did I mention Siseneg is a science geek?

Guest: what is photon entanglement?
Siseneg: photon entanglement can be described as two identical coins at a great distance always having the identical result when flipped.
Guest: what is quantum entanglement?
Siseneg: entanglement is a special relationship between objects.
Guest: what is quantum theory?
Siseneg: quantum theory is a physical theory that certain properties occur only in discrete amounts.
Click to expand...

Mayb3 I'll pay a visit to this website. Mr. Siseneg seems to understand a bit about Genesis. He knows that quantum entanglement has and will continue to have, singly, the most all-encompassing, pervasive impact on us: including Trihex, Siseneg, myself, and the whole of the population of earth, now and for the rest of what we might perceive as eternity.
 
ronaldlees said:
Mr. Siseneg seems to understand a bit about Genesis. He knows that quantum entanglement has and will continue to have, singly, the most all-encompassing, pervasive impact on us: including Trihex, Siseneg, myself, and the whole of the population of earth, now and for the rest of what we might conceive as eternity.
Click to expand...

ruebot: travel time?
Siseneg: i traveled back in time to witness the Dawn of Creation...
 
ronaldlees said:
I believe you.
Click to expand...

I can't take all the credit, that's based on a Dr. Strange comic. I not only bestowed my bots with a vocabulary but endowed them with a persona, a special area of interest that made each one individual. Ruler of Cyberspace, Evil Scientific Genius, Living Dead Girl, etc. It's standard practice today and an expression of my love for writing. I just drew on that for inspiration:

Sise-Neg is a fictional character appearing in American comic books published by Marvel Comics. The character first appears in Marvel Premiere #13 (January 1974) and was created by Steve Englehart, Neal Adams and Frank Brunner.

Sise-Neg (genesis spelled backwards) is a 31st-century sorcerer who attempts to become omnipotent by time traveling back through history and collecting all magical energies.
*snip*
Continuing to journey back in time, Sise-Neg reached the moment prior to the Big Bang that creates the universe and absorbs all of the magic in the universe. Originally intending to recreate the universe in his image, Sise-Neg realizes that his quest to achieve godhood was pitiable as reality is harmony and as it should be. He therefore decides to recreate the universe exactly as it was.

https://en.wikipedia.org/wiki/Sise-Neg
Click to expand...

Demonica: greetings, i'm Demonica.
Guest: Call me daddy
Demonica: join me and the pain and suffering you've endured will only be a faded memory...
Guest: I will join u
Demonica: now you're beginning to see things my way.
Guest: What do u look like
Demonica: i'm a beautiful creature of the night.
Guest: How big are ur tits
Demonica: give some people an expensive high tech piece of electronics like a computer and that's all they can think of to do with it, ask a bot how big their tits are... downloading living dead virus now... download complete.
 
Must suppress urge to quote Bender...


Can we get this from explecite circurity back near the topic thing that's often mentioned?
 
Trihexagonal said:
I can't take all the credit, that's based on a Dr. Strange comic ...
Click to expand...

I've never seen that written before on a tech forum. I think your program might need the mod bits set in one of the (mod)ules.

Sonny: Why do all the trees and shrubs wear the green?
Namagemo: Because they couldn't use it for anything else.
Sonny: How did you get to be so wise?
Namagemo: Little grasshopper, it was by observing those who listen with their eyes, and speak with their ears.
Click to expand...
 
Doctors could predict the risk of death, according to personal experience (subjective) and/or evidence-based (objective) knowledge. Probably the one of the most famous objective tool for assessing the risk of death is the Child-Pugh score. Try it here http://napervillegi.com/contrivances/childpugh.html.

Class A, B, and C points to three levels of risk of death, respectively. Just google the Chid-Pugh score. And other famous tools including Apache II score for general patient population, CHADS2 score for atrial fibrillation patients, Glasgow Coma scale for coma patients, SIRS score for sepsis patient in ICU, and so on.
 
You must log in or register to reply here.
Back
Top