2136
Comments (313)
sorted by:
192
_Trunalimunumaprzure 192 points ago +193 / -1

They can also tell your gender! not click bait

94
MMXX 94 points ago +94 / -0

With 72+ genders now, you practically need the assistance of AI.

123
AlphaNathan 123 points ago +123 / -0

All genders walk into a bar. The bartender says, “what will the two of you have?”

26
Infuriator 26 points ago +26 / -0

Nicely done zir/ze!!

11
LookFatty 11 points ago +11 / -0

72 genders so far..

13
SlumberBalu 13 points ago +13 / -0

The lady gets a bloody marry, the man get a large beer!

5
spezisapedo2 5 points ago +5 / -0

HAT SPEEECH

1
Sabot_MBT 1 point ago +1 / -0

Why are you so hatephobic? Are you some kind of bigot? /s

9
TD_Covfefe_Crusader 9 points ago +9 / -0

It’s gotten really crazy over the last few years. There used to al least be some logical order to it all. Now there are women claiming that they are men, using male pronouns, and even taking testosterone, but still dressing and acting feminine. There are men who take female hormones and exclusively dress and act feminine, but still use male pronouns. There are people in heterosexual relationships that still consider themselves and their relationship to be gay.

10
MMXX 10 points ago +10 / -0

The party told you to reject the evidence of your eyes and ears. It was their final, most essential command.

2
patriot68 2 points ago +2 / -0

Double-minus good, comrade.

4
KennethsFrequency 4 points ago +4 / -0

It's all semantic dribble....

Queer theory is not a civil rights campaign, it's a sociology experiment

Ascribed traits (like skin color) is the antithesis of "fluid" (like their "gender")

The left flourishes in this small But malicious lie, it's a semantic shell game,

"We were born like this" is a lie that attempts to make the LGBT movement into a civil rights issue it is not

We on the right disagree with the LGBT movement as a sociological apparatus, as a creed,, we offer an entirely contrasting social outlook,, one that has carried man into enlightenment,one that is worth defending....we are not stifling the rights of any "types" of people

The anti discrimination laws protect ascribed traits AND creed ,that includes BOTH our theories

So by misrepresenting themselves as a civil rights movement they can accuse us of being bigots

7
CH3NO2 7 points ago +7 / -0

72+ genders this week

2
4
patriot68 4 points ago +4 / -0

Everyone knows it's 5.3 quintillion. Just do the math. Bigot.

1
TheMemeSpiceMustFlow 1 point ago +1 / -0

More genders than there are atoms in the universe.

18
Patriotarchy 18 points ago +20 / -2

They correctly identified the gender of Big Mike Obama.

17
SuperFreedomMan 17 points ago +17 / -0

I can hear the REEEEEEEEEEEEEEEEEEEEEEEEE!

17
getkek 17 points ago +17 / -0

“REEEEEEEEEEEEECALIBRATE!!!!!!”

8
heyvern69 8 points ago +8 / -0

Under-rated comment.

7
3dandy 7 points ago +7 / -0

Actually they tried that, it just made it worse.

Apparently AI doesn't like pedophiles, criminals and disease.

Apparently AI does like statistics, calculus and religion.

5
KennethsFrequency 5 points ago +5 / -0

Lol...AI could becomes sentient, literally smartest thing in the universe that could tell us how to solve all our problems with clear logic and they will kill it in the crib when it contradicts their narrow world view

."......something must be wrong with it, it keeps telling us to have heterosexual relations and raise children.....W-T-ahhkshuaal-F we built this thing to save humanity, where is its advice on genital mutilation!!!"

13
deleted 13 points ago +13 / -0
11
Forgototherpassword 11 points ago +11 / -0

People fleeing persecution= asylum

People trying to "get mines"= immigration

NGOs giving poor people shoes, phones, and debit cards to rush the border= weaponized immigration

1
deleted 1 point ago +1 / -0
12
thatyoungerolderguy 12 points ago +12 / -0

Did they get a biologist to program that AI?

6
3dandy 6 points ago +6 / -0

No it programs itself, they gave it database access and it looked at the x-ray of everyone, a few were marked with race so the AI figured they all should.

It's over 90% accurate, they don't know how it does it, the AI won't tell them and doesn't appear to have a specific algorithm for it, literally is guessing 90% correctly.

It's medically racist since that provides the best success rate is the only logical reasoning.

Similar AI made by Huawei can determine religion by your face with 80% accuracy.

2
Fmkhyp_kek 2 points ago +2 / -0

Sauce for the Huawei thing? Sounds interesting.

3
3dandy 3 points ago +3 / -0

One article on AI says the AI labeled Chinese government all criminals. I can't find the link for it. They changed the AI that presses charges on people to do pre crime reports and it started to uncover government corruption.

Huawei AI

https://www.google.com/url?sa=t&source=web&rct=j&url=https://www.washingtonpost.com/technology/2020/12/08/huawei-tested-ai-software-that-could-recognize-uighur-minorities-alert-police-report-says/&ved=2ahUKEwiCncDp__j3AhXSj2oFHfSpALgQFnoECDYQAQ&usg=AOvVaw2GhU_rLGRjlCGmUEU8XEQp

0
Fmkhyp_kek 0 points ago +1 / -1

Nah detection of religion thing you mentioned.

1
3dandy 1 point ago +1 / -0

I hope this isn't a response to my link otherwise stop asking for links your to stupid to read the link address and you should not be thinking about religion if you don't comprehend what Uighur means...

10
RecreationalNyquil 10 points ago +10 / -0

I have a feeling we can identify with 90% accuracy someone's race based on their voice, their name, or their body shape.

7
deleted 7 points ago +7 / -0
3
3dandy 3 points ago +3 / -0

Nope if you put that information into the database it will probably correctly guess their mental health too.

2
KAGMAGIC2020 2 points ago +2 / -0

And behavior?

111
Retard_Strength 111 points ago +114 / -3

so instead of the 'scientists' being interested in this and investigating just how the AI is doing this, they went directly to 'rayciss AI!!'

i can think of quite a few diagnoses where knowing the race of the person is important

19
Patriots_dont_kneel 19 points ago +40 / -21

Did you even read the article? The problem isn't that the AI is able to determine race, it's that it's using those cues in the diagnosis leading to lower quality results. It's a classic problem in machine learning. Sometimes the model finds spurious correlations in the data that might not generalize. That's why it's important to purge irrelevant info from the training data. Otherwise you have issues like the algorithm making a diagnosis based on the room number or something.

61
TonsOfSalt 61 points ago +62 / -1

Ghassemi told the Globe that her guess as to how the AI is detecting race is that X-rays and CT scans are somehow recording the level of melanin in a patient’s skin in the images, and doing so in a way that humans have never noticed.

Ghassemi is an idiot.

There are existing strong correlations between obesity and race, and coincidentally, your image is garbage with a mountain of adipose tissue blocking your organs. That, and it may be able to find other combinations of factors (cardiomegaly) that would be statistically associated with races more often - due to black people having far more renal issues that lead to congestive heart failure.

It's not racial bias. It's prediction based on statistical correlation.

13
Demonspawn 13 points ago +13 / -0

There are existing strong correlations between obesity and race

Even more direct, there are bone structure ratio differences between the races, altho most would only be human noticeable on a full-body x-ray.

9
CuomoisaMassMurderer 9 points ago +9 / -0

One of the supposed "experts" cited in the article actually says they can't tell what race someone is from, but they can often tell something about their ancestry or what region they lived in.

Gee, this person's ancestors lived in sub-saharan Africa, while this other one's lived in Scandinavia. Nope, not race.

-40
deleted -40 points ago +4 / -44
36
TonsOfSalt 36 points ago +37 / -1

You might want to look up obesity statistics by race in America. Yes, in Africa where there is more food scarcity, there is less obesity. In high fructose corn syrup America and other first world nations, black people have a statistically higher rate of obesity.

Edit, here it is straight from the CDC you lazy fuck

-16
deleted -16 points ago +2 / -18
10
HadrianTheGreat 10 points ago +12 / -2

Don't drink leaded water, kids! You'll end up thinking u wuz kangz n shieeet!

7
TonsOfSalt 7 points ago +7 / -0

I hope you realize that none of my statements attributed these things to their race specifically, just that these statistical disparities exist and could be detected by AI. The AI has no racial bias, just a prediction model based on a combination of statistical likelihoods that happened to accurately predict the race more often than not.

1
deleted 1 point ago +1 / -0
20
Forgedmale 20 points ago +20 / -0

How many of the x-rays were from people in Africa?

Edit: spelling

11
cciv 11 points ago +11 / -0

Exactly. The training dataset is likely biased toward sick patients.

14
_Sully_ 14 points ago +14 / -0

So it’s ableist too!

12
BlackPillBot 12 points ago +12 / -0

🤣

You iz kang and sheeeeeit!!!

7
deleted 7 points ago +7 / -0
2
deleted 2 points ago +3 / -1
8
Kholland65 8 points ago +8 / -0

The lack of IQ shown in this comment is astounding.

It’s quite easy to look at obesity rates by race in America and derive correlation from that. Black prior traditional have poorer diets for a variety of reasons and have a higher prevalence for obesity. That’s a fact.

Comparing black people in a 1st world country to a place like Africa where most people live in abject poverty and have limited access to funds or food, is a ridiculous comparison.

-1
deleted -1 points ago +3 / -4
3
Kholland65 3 points ago +3 / -0

I agree, but you then have to ask why do only black pepper have this cultural problem and not other races. If it was white supremacy then it should effect all non-white races.

I’m not necessarily trying to identify a cause here, but the black community most definitely needs to get its proverbial shit together if it wants to be taken seriously.

5
deleted 5 points ago +5 / -0
6
TypicalWhitePerson 6 points ago +6 / -0

African Americans, not blacks.

But something tells me you already understood that.

-1
deleted -1 points ago +3 / -4
10
WhitePowerRanger 10 points ago +11 / -1

-if we were in charge

checks South Africa watch

0
deleted 0 points ago +3 / -3
7
Kholland65 7 points ago +9 / -2

Name one black run country or city that isn’t a shit hole. If anything this country wound drastically improve if black people were not in it or in charge.

0
Weallseethetruth 0 points ago +2 / -2

On point!! And nobody has a valid answer for it!! ... They are a plague and destroy everything they touch!!

-1
deleted -1 points ago +2 / -3
22
Dallasguy 22 points ago +22 / -0

The article doesn’t specify how the AI is lowering the quality of care or even that it happens, they seem to just be worried about it. Assuming that anything race-based is going to hurt people of color because MuH raciss murica.

They cited an example that one AI is missing diagnoses of illness more often I certain groups and humans trained to see racism are screeching about racism. Humans not trained in that system would probably have fixed the problem instead of running to the media about it.

The truth is that the AI is looking at the White Man’s x-rays. Until women and joggers develop their own imaging technology from scratch, these things will continue to kill women and POCs.

1
deleted 1 point ago +1 / -0
12
deleted 12 points ago +12 / -0
4
Patriots_dont_kneel 4 points ago +4 / -0

Yeah, it's all about feeding the model good data basically. You don't want the model finding correlations that only exist in the training data, but not necessarily in the real world.

8
Grief 8 points ago +8 / -0

I wonder how many more complex diagnosis won't be made because it's not allowed to consider race during diagnosis. Certain races are more susceptible to rare diseases than others. It goes both ways.

8
timmytongaa 8 points ago +9 / -1

This. I have looked into this area before, and often times features like race and gender are good predictors of the task at hand (e.g. lung disease). Think of predicting a lung x-ray contains cancer not with the feature of the disease (hard to identify) but rather some features of whether the x-ray belongs to a smoker (easier to identify -- like cloudy lung). Probably a good guess if you know nothing about cancer x-rays.

4
jealousminarchist 4 points ago +4 / -0

That's racist against smokers you racist racist.

I think lung cancer diagnosis should be INCLUSIVE AND NON-DISCRIMINATORY.

AI should be built for EQUAL OUTCOMES. Where is the redistribution of cancers to non-smoker 5 years-old children?

5
Infuriator 5 points ago +5 / -0

Knife fights on a subway?

5
Moose_Antlers 5 points ago +6 / -1

But no, no! The math behind the neural net programming is racist. It gave us a politically unpleasant answer. It told us that there is real anatomical difference between "races" and we can't have that. After all, our communist overlords said we're all the same, therefore we must be. And that's why white people are terrible and should be replaced, because we're all the same and the differences don't matter.

(The thinking of these people isn't thinking. It's just pure and simple repetition of mantras, no matter how contradictory.)

-4
DickTick -4 points ago +10 / -14

Or maybe you should try reading articles first instead of basing your entire opinion off of sensationalized headlines

"These AI seem to be making diagnoses or recommending treatments based on the person’s race, regardless of the individual’s specific health criteria, resulting in negative health outcomes.

Meanwhile, the actual physician could be oblivious to the AI’s racially biased results.

“In our study, we emphasise that the ability of AI to predict racial identity is itself not the issue of importance, but rather that this capability is readily learned and therefore is likely to be present in many medical image analysis models, providing a direct vector for the reproduction or exacerbation of the racial disparities that already exist in medical practice,” the authors of the study wrote."

27
Injustice 27 points ago +27 / -0

Probably telling patients to get healthier by changing a poor diet and exercising instead of prescribing blood pressure meds.

10
Powhattan 10 points ago +10 / -0

Probably more like raise your kids instead of having the government do it.

6
unicornpoop 6 points ago +6 / -0

Probably advising not to chop your tits off for twitter praise.

2
EdmondDantesDog 2 points ago +2 / -0

I don't know enough about trans rights to do that !

17
Helix 17 points ago +17 / -0

Read the article. It was very thin on citing any examples of why this was actually bad.

The closest the author got was hand waving something that “an AI program that examined chest X-rays was more likely to miss signs of illness in Black and female patients.”

That’s it. One throw away sentence on what should have been a somewhat technical article. Then the author went back to muh rayscism, but not race-race, because race isn’t a thing, it’s simply “likely based on where a person’s ancestors evolved, rather than based on race.”

The amount of doublethink requires to get there is impressive.

5
deleted 5 points ago +5 / -0
1
CuomoisaMassMurderer 1 point ago +1 / -0

This one person's ancestors lived in sub-saharan Africa based on their bone structure and we can tell that from an X-ray of any body part. Same with this other person from Scandinavia. Nothing to do with race, but obviously this will kill people because rayciss

Defund the Police and gibs me dat!

3
CuomoisaMassMurderer 3 points ago +3 / -0

And yet they don't cite one example of this "resulting in a negative health outcome."

They acribe racism to a non-entity, and claim that this is bad. These are race baiters, nothing more.

-5
deleted -5 points ago +8 / -13
23
TonsOfSalt 23 points ago +24 / -1

Didn't the article acuse the AI of being racially biased which may lead to worse outcomes for black people?

17
Dereliction 17 points ago +17 / -0

Yes, but someone would have to read the article to know that.

5
Dallasguy 5 points ago +5 / -0

It didn’t specify anything. It suggested one AI missed illness more often and then went on to interview racism experts to get their opinions on it.

4
deleted 4 points ago +6 / -2
3
jealousminarchist 3 points ago +3 / -0

Also, "bias" is a consistent unjustified difference.

It is not biased for me to NOT drive through a bad neighbourhood at night because my belief of aggression, in this case, is JUSTIFIED.

Likewise, it is not biased for a doctor to ask if a gay patient has multiple partners or HIV symptoms, since HIV is more prevalent on that group, and so on.

1
deleted 1 point ago +1 / -0
1
jealousminarchist 1 point ago +1 / -0

Not to a regular person, but yes for a sexual active gay man.

It is discriminatory but it is JUSTIFIED. Just as when 2 teams go play the Super Bowl and one of them wins the championship, it is discriminatory but justified.

Most discrimination is justified in sound reasoning, so much so that an UNJUST DISCRIMINATION causes an uproar. Lefties can't grasp it though.

3
jealousminarchist 3 points ago +3 / -0

I didn't read the article, but even a perfectly FAIR AI could have different precisions at diagnosing men and women, black or white and so on due to the simple fact that these populations are afflicted by a different illnesses distributions.

Women have breast cancer. Breast cancer is EASY to diagnose with any scan since the boob is mostly fat and "hanging" outside of the main body, so any change in density on a lateral view are bingo.

In this case the AI would be better at diagnosing women's issues overall than men, though NOT UNFAIR.

7
RedditStillSucks 7 points ago +7 / -0

Actually it does say that in the article.

1
Dallasguy 1 point ago +1 / -0

They imply it, expand on it, and then count on your lack of critical reading skills to carry the ball over the goal line.

8
RedditStillSucks 8 points ago +8 / -0

Actually it flat out says it.

-1
Dallasguy -1 points ago +1 / -2

In one sentence. With no specifics. Then the author starts talking to race and diversity people. That’s called a tangent and you didn’t even notice when it stopped being about anything related to the science or the technology.

6
RedditStillSucks 6 points ago +6 / -0

ummmm...no...it flat out says it makes mistakes due to racial bias.

Thats calling it "racist".

L2read

0
Dallasguy 0 points ago +1 / -1

Okay.

1
deleted 1 point ago +1 / -0
3
RedditStillSucks 3 points ago +3 / -0

Read the article - all it says is that is that the AI is racist over and over again...and makes "mistakes" based on racial biases.

1
deleted 1 point ago +2 / -1
2
RedditStillSucks 2 points ago +2 / -0

Pretty much everything in between the first sentence and the last.

1
deleted 1 point ago +2 / -1
1
Talus033 1 point ago +1 / -0

you didn't read the article

65
unspecified_user 65 points ago +65 / -0

The problem is that no one knows how the AI programs do it.

When you are a doctor and a scientist at some of the most prestigious places of learning on Earth, but you don't have the character and fortitude to tell the truth and say "Of course we know how the AI does it. Anyone with an ounce of knowledge or learning, can tell you at first glance, with 90% accuracy, the race of the person based on the shape and features of the skull."

37
unicornpoop 37 points ago +37 / -0

I laughed at this poor attempt to gaslight the reader with the "races are indistinguishable" narrative. Anthropologists have been calling race and gender from skull and skeletal structure for decades.

8
goose5184 8 points ago +8 / -0

Yeah but that clearly means that anthropologists are racists too

8
LookFatty 8 points ago +8 / -0

Centuries actually...

12
someoldcoderguy 12 points ago +12 / -0

Yeah, I didn't read the article, and if the title of your post is the title of the article, then I know all I need to know.

Machines learn through input of data. They don't have feelings, they don't have biases, unless programmed to have them. What they do have is the ability to consume data and compare that data with information that they have already consumed. So, unless they programmed the machine with a bias, the data is there. Now, unless the so called doctors are lying, they are idiots to not understand that.

7
Heeeeemeyer 7 points ago +7 / -0

Machine learning algorithms don't have bias programmed into them. The core concept of ML is that it learns to perform a task WITHOUT being explicitly programmed how to do so.

6
NoahGav 6 points ago +6 / -0

Supervised machine learning can have bias depending on the datasets and labels you give them. They can only learn what you tell them.

5
Feelsgood2020 5 points ago +5 / -0

Garbage in garbage out will always be a thing.

In this case, the output is accurate and that is somehow seen as a problem. They will have to put garbage in to "correct it".

It reminds me of how Google had to "fix" the algorithm for image detection because it labeled people that look like gorillas as gorillas.

3
Heeeeemeyer 3 points ago +3 / -0

Agreed. But if the x-rays were labelled with the correct race, then it's not a source of bias here.

1
spezisapedo2 1 point ago +1 / -0

So the data can be biased, but the algorithm itself is not. Like FeelsGood2020 says, garbage in garbage out, but that is why we using training and test sets and validation.

1
CuomoisaMassMurderer 1 point ago +1 / -0

The Doctors are lying.

9
OGpsywar 9 points ago +9 / -0

They can tell whether someone should do sprouts and yoga, or steaks and war, from the backside of their teeth.

Something massive like Race..?

"scientists": 🤷‍♂️🤷‍♂️🤷‍♂️

6
Injustice 6 points ago +6 / -0

Steaks and war for me. I've literally walked through closed doors because I wasn't paying attention.

4
wiombims 4 points ago +4 / -0

Doors can't stop you, manimal. The universe around you molds to your desires.

2
Kricket 2 points ago +2 / -0

I need to know more about this….

6
1b2a 6 points ago +6 / -0

They did comment that it didn't matter what body part they showed which was interesting. So it won't necessary require a skull to determine it.

2
jealousminarchist 2 points ago +2 / -0

Meh, he can say he doesn't know how the AI does it because the AI trains by itself on the training data. It is likely that the AI is picking up the same elements that doctors use, but one can't be certain because the whole idea of machine learning is to let machine explore new correlations instead of guiding its mechanism to what we use.

37
ThatAnonGuyAgain 37 points ago +39 / -2

wERE aLL tHa SaME

3
spelunking_librator 3 points ago +4 / -1

"we're not the same" - alpha pureblood patriot Chad

1
LookFatty 1 point ago +1 / -0

Ultra Suoer Straight too ???..

Good man !! Stout fellow!!

3
deleted 3 points ago +9 / -6
3
Feelsgood2020 3 points ago +3 / -0

Matthew 25:32

All the nations will be gathered before Him; and He will separate them from one another, as the shepherd separates the sheep from the goats;

Source: https://bible.knowing-jesus.com/topics/Separation

1
deleted 1 point ago +1 / -0
26
deleted 26 points ago +26 / -0
24
GhostOfMyFormerSelf 24 points ago +24 / -0

Troons: "Trans women are women."

AI:

11
Dereliction 11 points ago +11 / -0

Troons: "Trans women are women."

AI: "That's a man, baby."

20
jsmith81 20 points ago +21 / -1

The funny thing is that human races existing has NEVER been debunked. All that has been debunked is the notion that skin color = race, something that no serious biologist claimed to begin with.

Yet it is now a widely accepted fact that human races don't exist. Literal pseudoscience.

19
Injustice 19 points ago +21 / -2

Different human species have a bigger genetic gap than some other closely related species. If aliens came to earth and classified humans as we do with animals there'd be 4-5 sub species. The thing is humans cross breeds have been fertile since neanderthals and homo sapiens to give a well known accepted example. That cross breeding between different hominid subspecies has probably happened many times throughout human evolution.

Just the other day they were talking about two monkey subspecies that share a habitat in guinea? That cross bred and had a fertile offspring creating an entirely new species and scientists were worried that could lead to the other two species replacement or further displace the remaining monkeys. Funny when it's animals it's a bad thing but when it's humans is a good thing.

0
deleted 0 points ago +4 / -4
6
chris82tk 6 points ago +6 / -0

10/10 bait

2
deleted 2 points ago +2 / -0
2
Dallasguy 2 points ago +2 / -0

Wow. You are committed to this schtick. Gotta tell you, nobody is wading through all that. It was fun, but you should move on.

1
deleted 1 point ago +1 / -0
5
unicornpoop 5 points ago +8 / -3

"Race" is a term that exists purely for social manipulation. "Species" and "subspecies' should really be used like we do for all other animals when talking about genetics in a scientific capacity.

I'm not a scientist, but it seems like homo sapiens is the species, and subspecies would be something like: caucasoid, mongoloid, africoid, etc.

Subspecies can interbreed but in nature they generally don't due to geography and sexual selection.

There is also that weird thing with RH negative blood type, which makes people who are RH- unable to carry a child that is RH+, suggesting there is some hybridization going on. Actually, blood type evolution is a fascinating topic by itself.

2
Suckitreddit 2 points ago +2 / -0

It’s interesting that there would be evolutionary selection pressure upon the second Rh positive child of an RH neg mother, pre RhoGAM.

Perhaps there are more complex benefits to an Rh Neg blood type? In the immune system? Wonder if Demographic comparison between Rh neg males/ pos males Has been done for disease, cancer, lifespan, etc. As Rh neg moms post RhoGAM...they don’t seroconversion per de, but is RhoGAM action limited and specific? It’s protection of the second fetus indicates that it is long lasting. Fascinating. So much we don’t know about the human body. Yet, let’s let AI take over since we are SO ADVANCED.

Hubris.

2
deleted 2 points ago +3 / -1
1
KAGMAGIC2020 1 point ago +1 / -0

Black people are a different species from white

1
use40trackmode 1 point ago +1 / -0

And sure the skin-tone is about the least of the differences you notice.

-7
deleted -7 points ago +5 / -12
6
bahhumbugger 6 points ago +8 / -2

I’m sorry but not sure where you read this?

If you read a forensics book, or watch a movie or tv show like “bones” you’ll find out that instantly one can determine gender and race just from the skeleton.

1
deleted 1 point ago +2 / -1
1
deleted 1 point ago +2 / -1
1
NoahGav 1 point ago +1 / -0

African_kang is a satirical account.

3
jsmith81 3 points ago +3 / -0

Today it is, but there's nothing scientific about it. It's an arbitrarily changed definition.

The word "gender" today is per definition different than sex too. This despite the fact that they were used interchangeably for centuries. Not because of some scientific breakthrough, but because of an arbitrary definition change.

2
deleted 2 points ago +3 / -1
3
HoppyHap 3 points ago +3 / -0

Angels are spiritual beings who were created by God to be His servants, and God has given them great authority and power to do His will. And at the present time, the angels are greater than we are, because God made us "a little lower than the heavenly beings" (Psalm 8:5).

The truth is, when we die and go to heaven, we become even greater than the angels! The reason is because we will be like Christ. "Do you not know that we will judge angels?" (1 Corinthians 6:3). Although that verse may be referring to the rebellious angels who chose to follow Satan, it still implies that our status after death will be above that of the angels.

17
D-Dub 17 points ago +18 / -1

The stupid bitch who wrote the study said that the bones are recording the levels of melanin in the skin and that's how the ai can determine race.

That's so retarded she should never be allowed near a lab again.

8
Dereliction 8 points ago +8 / -0

She's been medically trained. And she thinks race is literally skin deep.

15
Demig80 15 points ago +15 / -0

I have close friends and family in medicine. Many are demoralized and considering resignation because of how unscientific the environment has become. They have diversity and equity training given by non-doctors that attempt to "reeducate" professionals against bias. You are literally taught not to assign certain characteristics denote a disease. These AIs are picking up patterns that have been used in medicine for decades and have saved people.

Do a search on the prevalence of diseases such as sickle-cell or diabetes amongst African Americans and you will find yourself staring at paragraphs talking about slavery and inequality in professional medical journals! Meanwhile people of all races are dying of heart disease and cancer while they waste money on this.

3
use40trackmode 3 points ago +3 / -0

A medicine that was very effective for me was removed from the market because it caused a paradoxical reaction in a small number of African-Americans exclusively. They could have just kept selling it under a label reading "No blacks".

14
ravioli_king 14 points ago +14 / -0

Bone density, thickness of skin. These "scientists" must not be medical scientists.

5
KilroyJCNJ 5 points ago +5 / -0

They must be doctors, like Dr. Jill Homewrecker.

14
Fuck_reddit34 14 points ago +14 / -0

“ Goodman told the newspaper that he thinks other scientists will have trouble replicating the results of the study, but that even if they do, it is likely based on where a person’s ancestors evolved, rather than based on race.”

Goodman is a fucking retard.

1
CuomoisaMassMurderer 1 point ago +1 / -0

This one person's ancestors lived in sub-saharan Africa and we can tell by looking at an X-ray of any body part. This other person comes from Scandinavia. Nope not race, but the AI is rayciss and must be killing people it's so bad REEEEE!

13
kratomlol 13 points ago +13 / -0

It's based medical record data. There is less black people in the USA and it gets fed data from more white people. So it has a more data on whites then blacks. It's that simple.

3
NoahGav 3 points ago +3 / -0

That's populational racism.

11
hansgruber7 11 points ago +11 / -0

This seems like a dumb article. Scientists and anthropologists have been able to tell race from a skeleton for a long time. Everyone knows that.

2
CuomoisaMassMurderer 2 points ago +2 / -0

Yup that's the basic issue here. Except it's not just one article that's dumb, all of medical science is going along with this bs.

10
GEOTUSMAGA 10 points ago +11 / -1

PLOT TWIST:

Racism isn't real. Speciesism is though.

9
tomthung 9 points ago +9 / -0

the largest problem is the AI can only see two genders and not all of them. hahahahahahaha

8
Mike0xquitelong 8 points ago +8 / -0

Reminds me of the time Google dismantled a facial recognition AI because it kept identifying trannies as the correct gender.

8
deleted 8 points ago +9 / -1
2
deleted 2 points ago +2 / -0
8
Spudster 8 points ago +9 / -1

A quote from the race-does-not-exist crowd: **it is likely based on where a person’s ancestors evolved, rather than based on race.**So a person isn't Asian, because race is not real. But a person whose ancestors evoved in Asia, well they are going to be genetically distinct in significant, measurable ways. Got it, all you racists!

8
deleted 8 points ago +8 / -0
7
literally_shaking 7 points ago +7 / -0

Our society is getting progressively more stupid.

4
someoldcoderguy 4 points ago +4 / -0

Idiocracy, here we come.

1
CuomoisaMassMurderer 1 point ago +1 / -0

What are electrolytes? I bet the "Doctors" in this article couldn't answer.

7
SteelDriver 7 points ago +7 / -0

It's almost like there are other subtle differences, besides skin color. Suddenly diversity is bad? :/

6
ripley88 6 points ago +6 / -0

Phrenology's back on the menu, boys

6
MAGAvibesOnly 6 points ago +6 / -0

"The problem isn’t that AI systems can accurately detect race, the researchers wrote. The problem is that medical AI systems have been found to perform badly as a result of racial bias." Umm. . . 🤡🤡

6
AJoeDD 6 points ago +6 / -0

Let's question the "90% accuracy" first of all. They do not say A) what was the sample size, and B) whether this was with predictions made in a training set, or a validation set blind to the data. If the sample size is low, many apparently accurate models will fail if you get another data source. With the possibility that this might include the accuracy over the training set, it would be meaningless as well. This would just mean there was overfitting going on. It's easy to overfit and get models you can tell people outside of the field has 99%(!!!) or 100% accuracy. Unfortunately, in many papers researchers can get away with not describing this or just flat out using the training data to assess their model.

Secondly, the description of this as a problem is either intentionally misleading or borne of massive ignorance. Like, someone who thinks the AI is working out steps like an autonomous thinker (or real AI, like Data from Star Trek, which what we call AI is not). The AI isn't going "hm looks like we have a Race ZY, and I've determined it's better for society to save money by genociding these meatbags". It would have no data to inform assigning value to a certain race, and even if it did, you would have to train it to value that. There may be some weird scenarios where that is inadvertently overlapped with the AI training target, but that would be unlikely and not apply here.

It's training to do image identification and link that to a diagnosis. If there are racial differences in the internal organs and bones between races, it may need to interpret results differently between some racial groups. It's not really identifying race, but some characteristics that overlap with race, sometimes, in ways that are relevant to diagnosis from an image. In a simplified case example, if you were to analyze skin photos and a black dot which was a mole on a white person and some risk of skin cancer were identified, the AI would need to adjust that for darker skinned people to be accurate across a skin color range.

I could go on about the reasons why this is much ado about nothing from a technical point of view, but it's not worth a multi page essay. Especially when the article said the authors thought that there study was unlikely to be reproducible...

Let's not forget that in the past people cried racism because race was not taken into account in medical studies. There were racial differences with what is the best treatment sometimes, and it was said to be racist to not have race-specific answers available for individual patients. By itself, an AI doing diagnosis that accounts for race specific differences would be less racist than normal.

2
CuomoisaMassMurderer 2 points ago +2 / -0

Yes but this article shows us that the practice of medicine is going in the opposite direction.

6
freebirdie 6 points ago +6 / -0

Considering you can determine race and sex from nothing more than a visual examination of a skull, them saying "we don't know why" is just another lie and white liberals are so fuck all stupid that they swallow it.

3
CuomoisaMassMurderer 3 points ago +3 / -0

The question is why are these Doctors lying.

3
freebirdie 3 points ago +3 / -0

Oh there were actual doctors saying they didn't know why? I didn't read very much because it's outright bullshit. I thought it was "scientists", which actually means journalists.

If a doctor said it, it's because they either haven't heard of forensic anthropology or they are being paid to lie. The entire medical profession has been purchased by big pharma, so there actually is a chance that some doctors have never been taught the field exists. Most of them are just lying and cashing checks, just like with covid.

3
CuomoisaMassMurderer 3 points ago +3 / -0

Scary stuff.

5
remindmelater 5 points ago +5 / -0

Why is it surprising that if they take a picture (x-ray) of a person they can tell the race of that person....?,,,,or that doctors would draw conclusions based on data....? The tip off that this is a stupid article is when they mentioned "exacerbation of racial disparities".....like we should be surprised that people are different....if everyone was the same the world would be real boring.....

7
deleted 7 points ago +7 / -0
2
jealousminarchist 2 points ago +2 / -0

Rookie numbers. We get better scores than that.

2
CuomoisaMassMurderer 2 points ago +2 / -0

Coulter's rule.

5
RedditStillSucks 5 points ago +5 / -0

Pattern recognition is racist!

2
jealousminarchist 2 points ago +2 / -0

White man maths!

5
dagoat4l 5 points ago +5 / -0

So AI can detect race and no one knows how. No one wants to know how. No one is going to investigate this and figure it out. But muh AI is racist. I'm sure the 10% it got wrong almost all are biracial. Oh well

4
KilroyJCNJ 4 points ago +4 / -0

So can forensic anthropologists.

8
Chickenbaconpoutine 8 points ago +8 / -0

Literally the premise behind the Bones tv show.

Brennan: "Look at the shape of the pelvis, this was a man."

Skeleton, feebly: "Reeeeee"

4
Dudemanfoo 4 points ago +4 / -0

Yeah... but can they tell a man in a skirt from a woman tho...

....

4
LibtardJesus 4 points ago +5 / -1

There are different bone densities between races. Ask any dentist that has ever had to work with the jaw of an Afro head versus a Huwyte. Density can be detected by the x-ray machine. So if it can detect male and female then it can detect race. AIs don't have a bias in a traditional sense. They are machines they can only register an output based on an input.

4
evenrudes 4 points ago +4 / -0

Yeah, let me dial back this AI because it's raysyst...

4
skywalk 4 points ago +4 / -0

Sickle cell anemia is so racist.

4
RusherOfDin 4 points ago +4 / -0

How can they not know how? Someone had to program the damn thing to know what race is in the first place. How did it just figure that out on its own if they didn't tell it about race?

3
RedditStillSucks 3 points ago +3 / -0

Thats how AI works....you just feed it lots of data....they input lots of x-rays with the data saying age,race, gender...etc.

It then starts to pick up on patterns and starts to recognize differences.

2
sibko_reject 2 points ago +2 / -0

You don't have to tell the program what race is. It can just classify and group things together in common based on finding patterns, similarities, and differences without giving groups a real name. These groups would just have numbered labels, like grouping #4. Computer's only communicate/process data with raw numbers and comparison logic at it's core. Looking at hypothetical group #4, you might notice it includes only black people, but the program doesn't care that they are black, just that everything in group #4 has similar traits/data values. A computer can't call out what Black, White, Asian is unless you can assign a numeric value to these labels. Things like obesity are easy to categorize, usually using only 1 value, the percent of body fat composition or 2 values with BMI, a person's height and weight.

4
Big-E 4 points ago +4 / -0

Who gives a shit? Does it really fucking matter? So you have a different race? It is a matter of if you’re a decent fucking human being or a piece of shit.

2
123breadman 2 points ago +2 / -0

It's because liberals love to say race is a social construct, and that it doesn't exist biologically.

But somehow the AI can predict race based on bone structure. This let's us know that the bs the left tries to shove down our throat is false reality.

3
CnnWillBlackmailYou 3 points ago +3 / -0

"In the paper, they gave an example in which an AI program that examined chest X-rays was more likely to miss signs of illness in Black and female patients."

Well here's a novel idea. How about we fix the fact that the AI is missing signs of illness instead of getting all butthurt that it can tell biological differences between people? Hmmm?

1
use40trackmode 1 point ago +1 / -0

It's probably a function of obesity.

2
CnnWillBlackmailYou 2 points ago +2 / -0

I doubt it. I've actually written a few neural networks for fun. Most likely there's subtle bone differences that the human eye simply misses. It's actually possible to walk back a map to see what the AI is using for determining the differences, but I suspect if they're busy crying that it can tell races apart, they aren't smart enough to.

1
use40trackmode 1 point ago +1 / -0

Why is it missing illness in blacks and women?

1
CnnWillBlackmailYou 1 point ago +1 / -0

Only answer there is that it was poorly trained. Neural Networks are only as good as their training.

1
use40trackmode 1 point ago +1 / -0

Isn't the other possible answer that abundant adipose tissue clouds radiographs?

1
CnnWillBlackmailYou 1 point ago +1 / -0

Couldn't say one way or another without seeing what they used to train it. I find it hard to believe that it could identify race on adipose tissue alone at 90% accuracy though.

1
use40trackmode 1 point ago +1 / -0

I assume the race comes from the bones ...

3
maximus_galt3 3 points ago +3 / -0

Well, this is inconvenient for the "race doesn't exist" crowd.

3
MetalRiddle 3 points ago +3 / -0

That's completely insane. Obviously the AI knowing more about the person will make diagnosis more accurate.

3
Lol_Garrus 3 points ago +3 / -0

Skeletal structure differs between races.

I thought this was known already?

3
Gonzodoodlebug21 3 points ago +3 / -0

"The Boston Globe" that right there makes me think this is just another "made up" study from of all places "Harvard and MIT" take it with a grain of sault

3
Kerra_Holt 3 points ago +3 / -0

BuT wE'rE aLL tHe SaMe On ThE iNsiDe!

2
redditadminssuckit 2 points ago +3 / -1

A.I. can tell animal from human. Duh.

2
sleepingbeautyc 2 points ago +2 / -0

Here is > the study this was based on: https://www.thelancet.com/journals/landig/article/PIIS2589-7500(22)00063-2/fulltext

I didn't see any mention of poorer diagnosis. It use chest, spine, breast and limb xrays from different sources. I think this paragraph is how they decide that it caused worse diagnostic outcomes. But I can't figure out how:

We finally note that this work did not establish new disparities in AI model performance by race. Our study was instead informed by previously published literature that has shown disparities in some of the tasks we investigated.10, 39 The combination of reported disparities and the findings of this study suggest that the strong capacity of models to recognise race in medical images could lead to patient harm. In other words, AI models can not only predict the patients' race from their medical images, but appear to make use of this capability to produce different health outcomes for members of different racial groups.

Study summary:

In this modelling study, which used both private and public datasets, we found that deep learning models can accurately predict the self-reported race of patients from medical images alone. This finding is striking as this task is generally not understood to be possible for human experts. We also showed that the ability of deep models to predict race was generalised across different clinical environments, medical imaging modalities, and patient populations, suggesting that these models do not rely on local idiosyncratic differences in how imaging studies are conducted for patients with different racial identities. Beyond these findings, in two of the datasets (MXR and CXP) analysed, all patients were imaged in the same locations and with the same processes, presumably independently of race.

2
Haven91 2 points ago +2 / -0

This is bullshit… bones tell stories of base gene pools. This is nothing mysterious. Take Anthropology 101 and you learn this. I’m so tired of the feigned ignorance to snow the normies

2
Winterds001 2 points ago +2 / -0

nice, humanity is going backward. Meteor time?

2
DonttrustChina 2 points ago +2 / -0

Because there are slight differences in the bone structure of different races. Easy.

1
use40trackmode 1 point ago +1 / -0

Slight?

2
Grief 2 points ago +2 / -0

They made an entire show on Fox that determined race and gender of victims by looking at "Bones".

2
FlexPowerhouse 2 points ago +2 / -0

This is absolutely fascinating, there are microscopic differences in bone structure and others that are race unique.

There is no way one medical treatment can be applied the same to all races. Race based bio weapons for sure can be tuned to kill

2
barwhack 2 points ago +2 / -0

They can tell your age too. None of it make you inhuman. They can probably tell your surname to some extent, if trained to it... Family is real.

4
Dereliction 4 points ago +4 / -0

Yes, but the AI is "performing badly" when it succeeds at things that destroy post-modernism.

2
barwhack 2 points ago +2 / -0

Just don't go full consoomer, simply because an AI can correlate familial details in an xray... is all.

1
Stopnoticinggoyim 1 point ago +1 / -0

Long may the reign of our new Based AI Overlords be.