Google AI now predicting if you'll die within 95% accuracy

[captain_to_fire]

11 minutes ago, GoldenLag said:

95% accuracy is terrible in the medical sence. 95% accurate tests om a mass scale is bad. Which are some of the reasons doctors dont recommend healthy people with no family history to go get a mammogram or certain other tests that can come out with a false-positive

 

Still quite cool though

They only trained their ML model from data from two hospitals that spans either four or seven years and they used data from old people ages 55, 56 and 57 which is not a diverse set of data. https://www.nature.com/articles/s41746-018-0029-1/tables/1

[GoldenLag]

2 minutes ago, captain_to_fire said:

They only trained their ML model from data from two hospitals that spans either four or seven years and they used data from old people ages 55, 56 and 57 which is not a diverse set of data. https://www.nature.com/articles/s41746-018-0029-1/tables/1

Still cool tech that has potential.

 

But it indeed needs a larger dataset to be even considered for commercial use. 

[captain_to_fire]

7 minutes ago, GoldenLag said:

They cant tell tye difference between cancers

Well some cancers are treatable with a high chance of remission if detected early. Also that is quite wrong to say that hospitals can't differentiate different types of cancers. Depending on the hospital and their equipment, many nowadays use genomic sequencing in conjunction with the typical imaging, microscopy and serology in diagnosing and even determining the right treatment.

2 minutes ago, GoldenLag said:

Still cool tech that has potential.

 

But it indeed needs a larger dataset to be even considered for commercial use. 

It's kinda like the gaydar ML with a small data.

I think a ML model that has been trained to a limited and small dataset should not be used commercially or else you will run into problems like false positives or false negatives.

[Dissitesuxba11s]

I was expecting this meme to be the notification:

[Brooksie359]

2 hours ago, Princess Cadence said:

Click-bait title, learned with the finest; LTT.

 

Seriously, fine medics can out of their experience and knowledge make as much as accurate prediction after your entry trial at the hospital, this just automates the process.

I mean if you take the title at face value then it is pretty underwhelming. I mean I can tell you if someone is going to die with 100% accuracy because the answer is always yes. 

[LAwLz]

1 hour ago, Princess Cadence said:

Are you serious? okay, forget people who do 5 years of college and another couple years of residency and later another many years of specialization are incapable of prediction on their area of expertise, we totally need AI or we're doomed.

 

Common Sense works fine, finding specific nonexistent research of every logic thing in life just to please someone's need to debate/argue is not practical in life.

Yes I am serious. I want you to provide me evidence that doctors can already do this type of thing manually, with a higher accuracy. That is what you claimed.

The strength of this technology is that it can analyze thousands upon thousands of different data points as well as cross reference it with similar cases.

 

Doctors already use similar techniques (through automation) but it is very costy and time consuming. Because of that they often limit themselves to a very limited number of variables when making predictions.

Right now, over 80% of all alarms in hospitals are false positives caused by, among other things, inaccurate alarm settings set by doctors and nurses. False negatives are also not too uncommon, for example 12.5% of patience with severe sepsis are not diagnosed properly because the criteria do not take enough data points into consideration.

 

The idea that a doctor can go through thousands of hand written notes, and make precise predictions based on cross referencing such notes would be laughable if it wasn't such a serious subject.

 

 

No, I am not saying we are all doomed without AI, but you are extremely ignorant about this subject and what potential it can have.

 

Being able to feed a neural net millions of patient logs to make accurate predictions for future health issues is not something humans can do through manual work. It's just not possible, no matter how many years of experience and education a person has.

 

 

1 hour ago, captain_to_fire said:

No citation needed for that. Medics, nurses and doctors are trained not only to save lives but to determine prognosis and chances or survival. For every minute that passes by, a stroke patient's survival is getting slimmer. So I don't know what kind of citation do you need? Medical records? Not gonna happen as those are strictly confidential. How about textbooks in emergency medicine?

All I heard was "I have zero evidence but I assume it is true".

Just as a reminder, Cadence did not just say doctors can make predictions too. She said that they can make as accurate predictions, using only the data from the entry trial.

I am not denying that doctors can make predictions. What I am asking for evidence for is that they can make it as accurately using only a limited amount of data. Now that she has doubled down on her claims, I will also argue against her point that this technology is not necessary and that we should just trust "common sense" when making predictions that could potentially kill someone.

[Shammikit]

what a time to be "alive"

[Brooksie359]

1 hour ago, GoldenLag said:

Which is why cancer tests arent tye greatest.

 

They cant tell tye difference between cancers.

 

Some grow so fast you cant be saved.

Some grow so slow they can kill you before you grow old.

Some stop growing for no good reason

Then you have the one that you can be saved from.

 

They treat all of them the same just to be shure. And the cancer treatments take a toll om the body.

 

Id still recommend taking the treatment, just dont go around testing yourself 24/7 unless the doctor actually recommends it

I mean some of the cancers grow so fast that they are really easy to treat. The cancer cells stand out quite a bit because they grow at ridiculous rate and allow for treatments they target fast splitting cells really affective. Actually saved my friends life that the type of cancer was super aggressive to the point where if it doesn't kill you right away it is super treatable. 

[Deus Voltage]

1 minute ago, Brooksie359 said:

I mean some of the cancers grow so fast that they are really easy to treat. The cancer cells stand out quite a bit because they grow at ridiculous rate and allow for treatments they target fast splitting cells really affective. Actually saved my friends life that the type of cancer was super aggressive to the point where if it doesn't kill you right away it is super treatable. 

Certain cancer is too aggressive to be treated though. I've been following someone with cancer on YouTube PeeWeeToms, and he suffers from a very rare type of aggressive cancer which no solution as of yet yielded positive results for.

[Brooksie359]

2 minutes ago, Deus Voltage said:

Certain cancer is too aggressive to be treated though. I've been following someone with cancer on YouTube PeeWeeToms, and he suffers from a very rare type of aggressive cancer which no solution as of yet yielded positive results for.

The type of cancer my friend had literally grew to the size of a football in the matter of days. I would say it is probably one of the most aggressive out there. 

[Deus Voltage]

1 minute ago, Brooksie359 said:

The type of cancer my friend had literally grew to the size of a football in the matter of days. I would say it is probably one of the most aggressive out there. 

I wish your friend a quick and smooth recovery, cancer has taken  too many lives already. Treating it (however vague that word may be nowadays) will bring a healthy amount of hope to all of us (especially those with not so healthy lifestyles, but we have ourselves to blame) 

[Eduard the weeb]

kinda S P O O K Y however if it helps the hospitals save patients lives then I guess its cool.

 

however would have been cooler to find out if I'm gonna die in a war trump declares on someone.

[Shreyas1]

* Looks at title *

 

Google: There is a 5% chance that you will be immortal!

[laminutederire]

2 hours ago, rcmaehl said:

It's predicting when you'll regularly die... It's predicting if you'll die within a period of time of being admitted to a hospital for an issue. Basically, would help doctors put a rush on things if someone is going to die extremely soon.

Yeah or  not treat people they deem to be dead already...

[Princess Luna]

26 minutes ago, LAwLz said:

No, I am not saying we are all doomed without AI, but you are extremely ignorant about this subject and what potential it can have.

I beg your pardon? I'm extremely ignorant? based on your miss interpretation of what I wrote in the first place? The prediction alone is not surprising it is expected for a well build AI, thus why I mentioned the title was a bit miss leading, you call me an ignorant for not realizing the potential it has on system of information when my reply was exactly saying the true benefit of it is in the automation of the process.

 

 I also said that medics are capable of prediction since that's their job and this is common sense without need of specific research to prove that's what they do for a living lolz... I love how you twist other people's word wrongly to try justifying yourself.

 

I thought Sweden folk had top class education, the more you learn about the world

[Brooksie359]

6 minutes ago, Deus Voltage said:

I wish your friend a quick and smooth recovery, cancer has taken  too many lives already. Treating it (however vague that word may be nowadays) will bring a healthy amount of hope to all of us (especially those with not so healthy lifestyles, but we have ourselves to blame) 

Yeah he has been off of treatment for a couple of years now. Still had to go through alot of treatment to make extra sure that it didn't come back. 

[LAwLz]

1 hour ago, Cookybiscuit said:

Probably so socialist countries with "free" healthcare can simply let people die on the operating table because Google said they'd be unlikely to survive.

Nope, that will not be the case. At least not in Sweden.

We continue to treat people we are 100% sure will die, even today.

 

 

 

 

 

For those wondering here are some quick stats. The reason why they are two numbers in some points is because they used two different hospitals. So it might have been 93% accurate at one of them, and 95% accurate at the other.

• The number of patience they analyzed were 114,003.
• They used an average of 216,744 discrete data points per patient when making the predictions.
• The predictions were 93% (vs 85%) and 95% (vs 86%) accurate for the 24h after admission predictions, compared to the standard method used at the hospitals today.
• The false positives for patients the hospital had predicted to be at high risk of dying, were half in the AI predictions. That is to say, if the hospital made 100 wrong predictions about patients going to die, the AI correctly predicted that 50 of the would live (while still saying the other 50 would die, which they didn't).
• The AI was between 24 and 48 hours ahead of the medical team when it came to prediction accuracy.
• The readmissions within 30 days accuracy was 77% (vs 70%) and 76% (vs 68%).
• For predicting long length of stay after 24 hours, the AI's accuracy was 86% (vs 76%) and 85% (vs 74%).

 

So no, doctors are not as accurate, and they are much slower.

[LAwLz]

27 minutes ago, Princess Cadence said:

I beg your pardon? I'm extremely ignorant? based on your miss interpretation of what I wrote in the first place? The prediction alone is not surprising it is expected for a well build AI, thus why I mentioned the title was a bit miss leading, you call me an ignorant for not realizing the potential it has on system of information when my reply was exactly saying the true benefit of it is in the automation of the process.

 

 I also said that medics are capable of prediction since that's their job and this is common sense without need of specific research to prove that's what they do for a living lolz... I love how you twist other people's word wrongly to try justifying yourself.

 

I thought Sweden folk had top class education, the more you learn about the world

How was that a misinterpretation? You even doubled down on your previous post.

 

Here is what you said, and how I interpret it:

 

3 hours ago, Princess Cadence said:

Seriously, fine medics can out of their experience and knowledge make as much as accurate prediction after your entry trial at the hospital, this just automates the process.

Here you're saying that medics can make as accurate predictions after having done an entry trial.

Not sure what "entry trial" means but I assume you mean the first set of tests and diagnose.

This is wrong, because the predictions from the medics are quite a lot less accurate, as I showed above.

Even if your point was that medics can make somewhat accurate predictions, I don't think that's a valid argument for why this isn't great news. The accuracy is far higher with the AI in this test.

 

All I did was ask for a citation on your claim, and you blew up and was very rude towards me. Making strawman arguments, going "are you serious?" and insinuating that I lack common sense, as you can see here:

2 hours ago, Princess Cadence said:

Are you serious? okay, forget people who do 5 years of college and another couple years of residency and later another many years of specialization are incapable of prediction on their area of expertise, we totally need AI or we're doomed.

 

Common Sense works fine, finding specific nonexistent research of every logic thing in life just to please someone's need to debate/argue is not practical in life.

 

If your point was "doctors can make predictions too" then... duhh...

I have 0 years of medical education but I can make predictions too. The ability to make predictions doesn't really matter though. What matters is the accuracy and speed, and saying that "this just automates the process" is a serious underestimation of the technology. It does is much faster than doctors, with much higher accuracy too.

 

 

 

 

29 minutes ago, laminutederire said:

Yeah or  not treat people they deem to be dead already...

Again, in Sweden we treat patients who we know will die.

We can already tell when someone is going to die in a lot of situations, but those people get treated anyway.

[Trixanity]

It's not that doctors can't do what the AI does but that doctors don't have the time to analyze and triple check everything hence a machine doing that work will be extremely helpful and can do a better job.

If you put experienced doctors to the task of doing nothing but reading data and analyzing it then sure the human expertise may win but doctors need to do more than that. And when they're so busy they'll slip up or miss something and even at the best of times the human factor can be detrimental; you can't know everything and you're not always at the top of your game. That's how you get news stories that patients that could have been saved died instead. It's mostly good enough though hence people aren't dying left and right but the prognosis of each patient might be better if a machine within seconds can correctly identify critical points that need to be addressed. That would improve health care overall. 

 

Essentially I agree with Lawlz on this one. AI can do better than experienced doctors because analyzing data is something machines can do quickly and accurately. That doesn't mean doctors aren't good enough although like any other profession quality will vary. This could raise the standard.

[Guest]

5 hours ago, Minibois said:

I thought this was gonna be one of those early 2000's website that will say: "You will die on xx-xx date in xxxx year!" that used to be spooky for elementary school kiddos like myself.

But no, this is actually useful, but only for hospital patients..

We're all going to die in 2012 anyway. 

[laminutederire]

8 hours ago, LAwLz said:

Again, in Sweden we treat patients who we know will die.

We can already tell when someone is going to die in a lot of situations, but those people get treated anyway.

In Europe as a whole to be fair.

I meant more that in countries like the US, it's totally what will happen since they mostly only have economics in mind for health matters.

[Misanthrope]

Talk about clickbait: Not you OP you're just quoting the article but come on: Guessing when you'll die is a hell of a lot more impressive before you add the crucial part of getting admitted to a fucking hospital which it's an extremely narrow set of circumstances.

[Guest]

Someone should probably tell Google that it’s easy to tell if someone will live or die really. 

11 hours ago, mynameisjuan said:

You mean like the accuracy doctors give their patients and family. "You have 2 months, 3 top"....."Four years later"

Cancer is really hard to predict. Although if you are getting Chemo then it’s inevitable. 

 

There are many diseases that are just an estimate of the average life expectancy and you have to take into account how weird the human body is. Take Stephen Hawking for example. 

[mr moose]

I have a doctors appointment this afternoon for a lump in my back, I think I'll cancel that appointment because it appears there are more experts in LTT and the forums are free.

 

Have I made the right decision? 

[Guest]

On 6/20/2018 at 3:38 PM, mr moose said:

I have a doctors appointment this afternoon for a lump in my back, I think I'll cancel that appointment because it appears there are more experts in LTT and the forums are free.

 

Have I made the right decision? 

Let me make my diagnosis. 

Is is hard? If so, go to a doctor. 

Is it soft? If so, go to a doctor. 

Is it red? If so, go to a doctor.

Is it brown? If so, go to a doctor. 

 

I'm really good.