Skip to content
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(PC Magazine)   Twitter is looking into a possible racial bias that's causing the photo preview algorithm to show white faces more frequently than black faces. Now if only it would eliminate orange faces   (pcmag.com) divider line
    More: Interesting, Barack Obama, Democratic Party, African American, United States Senate, Mitch McConnell, Republican Party, Bill Clinton, United States presidential election, 2008  
•       •       •

895 clicks; posted to Main » on 21 Sep 2020 at 9:54 PM (5 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook



44 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest
 
2020-09-21 8:37:28 PM  
Saw this earlier on Twitter with lenny and Carl used as an example  .
 
2020-09-21 9:55:57 PM  
Shutdown Twitter.  Problem solved.
 
2020-09-21 10:00:47 PM  

AmbassadorBooze: Shutdown Twitter.  Problem solved.


I've been saying this since they started. It's really telling when people think they can have intelligent discourse in a medium that limits you to 140 characters.
 
2020-09-21 10:02:21 PM  
I'd pay extra for a freckle filter.
 
2020-09-21 10:04:32 PM  
Michael Jackson said it doesn't matter. Obviously Twitter is bias against white women
 
2020-09-21 10:11:07 PM  

tasteme: Michael Jackson said it doesn't matter. Obviously Twitter is bias against white women


Yeah, yeah, yeah.
 
2020-09-21 10:13:44 PM  
i.pinimg.comView Full Size
 
2020-09-21 10:17:34 PM  

Jesus McSordid: [i.pinimg.com image 550x309]


came for this
 
2020-09-21 10:19:18 PM  
This thread is going to be a real shiat show. Please make sure your dogs are wearing protective ear equipment before continuing down this thread.
 
2020-09-21 10:23:01 PM  

Kalyco Jack: AmbassadorBooze: Shutdown Twitter.  Problem solved.

I've been saying this since they started. It's really telling when people think they can have intelligent discourse in a medium that limits you to 140 characters.


... and, yet, science Twitter has been enormously useful to me professionally and intellectually. I don't think I follow anyone, except MC Hammer, without a post-graduate education.

It turns out you can do quite a lot with short messages. Also, we do this new thing now on the internet where you click the text and it takes you somewhere else. You should check it out.
 
2020-09-21 10:23:53 PM  
It also prefers boobs.
 
2020-09-21 10:24:07 PM  

Pichu0102: This thread is going to be a real shiat show. Please make sure your dogs are wearing protective ear equipment before continuing down this thread.


Oh my god. I just realized something. Everyone but you is a racist!
 
2020-09-21 10:24:41 PM  
Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."
 
2020-09-21 10:25:18 PM  

Nurglitch: It also prefers boobs.


No, that's Titter.
 
2020-09-21 10:25:21 PM  

Nurglitch: It also prefers boobs.


White boobs
 
2020-09-21 10:26:05 PM  

The Pope of Manwich Village: tasteme: Michael Jackson said it doesn't matter. Obviously Twitter is bias against white women

Yeah, yeah, yeah.


You're singing it wrong. It's yeah, he, he, aow!
Kids these days, jeez
 
2020-09-21 10:30:45 PM  

Nurglitch: It also prefers boobs.


So it works perfectly?
 
2020-09-21 10:33:40 PM  

Pichu0102: This thread is going to be a real shiat show. Please make sure your dogs are wearing protective ear equipment before continuing down this thread.


Certainly you wouldn't suggest that the main page is a stronghold of the Kentucky Klan, filled with subliterate mouthbreathers?

Because that would be accurate.
 
2020-09-21 10:34:17 PM  

Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."


Programmers for years have been saying two very basic things, both of which get tossed out in the crush to call them all racists;
-
1) Sensors don't take pictures of objects. They record light. Specifically the number of photons bouncing off of objects. Objects that bounce fewer photons leave the sensors with much smaller sample sizes to try to interpret, and much smaller deltas to compare. Which leads to;
-
2) Computers don't see people. They mainly use tools like contrast detection to try to identify general shape.
This is harder to do if there reduced contrast between lighted and shadowed parts. This is as true of less reflective objects as it is less reflective people.
-
/In short photons are racist and we should boycott the universe until it comes up with more racially sensitive energy emissions
 
2020-09-21 10:40:46 PM  

Dryad: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

Programmers for years have been saying two very basic things, both of which get tossed out in the crush to call them all racists;
-
1) Sensors don't take pictures of objects. They record light. Specifically the number of photons bouncing off of objects. Objects that bounce fewer photons leave the sensors with much smaller sample sizes to try to interpret, and much smaller deltas to compare. Which leads to;
-
2) Computers don't see people. They mainly use tools like contrast detection to try to identify general shape.
This is harder to do if there reduced contrast between lighted and shadowed parts. This is as true of less reflective objects as it is less reflective people.
-
/In short photons are racist and we should boycott the universe until it comes up with more racially sensitive energy emissions


The point is that the algorithm, though not actually "racist", has a disparate racial impact.

No one is saying boycott anything. They're saying, hey, this is happening, and it would be better if it didn't. If you can't handle that, then make sure to stay out of the sun, snowflake.
 
2020-09-21 10:46:01 PM  

thurstonxhowell: Kalyco Jack: AmbassadorBooze: Shutdown Twitter.  Problem solved.

I've been saying this since they started. It's really telling when people think they can have intelligent discourse in a medium that limits you to 140 characters.

... and, yet, science Twitter has been enormously useful to me professionally and intellectually. I don't think I follow anyone, except MC Hammer, without a post-graduate education.

It turns out you can do quite a lot with short messages. Also, we do this new thing now on the internet where you click the text and it takes you somewhere else. You should check it out.


Okay I admit I only use twitter to follow hot celebrities and porn stars.

I was in college when it came out. I took one look at it and said this looks like a cross between a mens' room wall and one of those old beater cars you see where the entire back is cover in bumper stickers. Hard pass. Its association with the sentient Cheeto only served to bolster my original appraisal.
 
2020-09-21 10:46:53 PM  

tasteme: Michael Jackson said it doesn't matter. Obviously Twitter is bias against white women


I'm sure to him, it didn't.

People tend not to care if you're black or white if you've got $100 million in the bank.
 
2020-09-21 10:48:51 PM  

thurstonxhowell: Dryad: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

Programmers for years have been saying two very basic things, both of which get tossed out in the crush to call them all racists;
-
1) Sensors don't take pictures of objects. They record light. Specifically the number of photons bouncing off of objects. Objects that bounce fewer photons leave the sensors with much smaller sample sizes to try to interpret, and much smaller deltas to compare. Which leads to;
-
2) Computers don't see people. They mainly use tools like contrast detection to try to identify general shape.
This is harder to do if there reduced contrast between lighted and shadowed parts. This is as true of less reflective objects as it is less reflective people.
-
/In short photons are racist and we should boycott the universe until it comes up with more racially sensitive energy emissions

The point is that the algorithm, though not actually "racist", has a disparate racial impact.

No one is saying boycott anything. They're saying, hey, this is happening, and it would be better if it didn't. If you can't handle that, then make sure to stay out of the sun, snowflake.


Oh, I don't think anyone disputes it has a disparate racial impact. Or that it needs refinement.
But for ten years now any discussion about the technical limitations involved and how to minimize them gets buried under conspiracy theories, accusations, and racial grandstanding, finally to be torpedoed because any attempts to fix it will never achieve 100% racial parity therefore are unacceptable.
Because no matter how good it gets, it won't have racial parity. Because physics doesn't give a shiat about racial parity.
Its the reason I can stand in the sun for far longer than you and still not get melanoma the size of grapefruit, but on the flip side I am never going to show up on facial recognition as often.
-
/Actually, I consider that a win on both counts
 
2020-09-21 11:01:39 PM  

Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."


I strongly suspect it was nothing of the sort.  It is kind of a PITA to process a photograph of dark ANYTHING.  Go take a picture of a black lab next to a yellow lab and see which one is easier to process to detect all of the detail.  This is just simple reality.

That the results had a color bias has nothing to do with "racism."  They are going to have the same problem with someone blond and blue eyed tanned to a burnt crisp.  Is is just a side effect of the physics of the visible light spectrum and how we tune our images to the peak reception of our eyes.

Now, should Twitter work on improving their algorithms to compensate for basic physics?  Yes.  But to blame it on "racism" shows just how divorced from reality some people are.
 
2020-09-21 11:06:47 PM  

nobody11155: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

I strongly suspect it was nothing of the sort.  It is kind of a PITA to process a photograph of dark ANYTHING.  Go take a picture of a black lab next to a yellow lab and see which one is easier to process to detect all of the detail.  This is just simple reality.

That the results had a color bias has nothing to do with "racism."  They are going to have the same problem with someone blond and blue eyed tanned to a burnt crisp.  Is is just a side effect of the physics of the visible light spectrum and how we tune our images to the peak reception of our eyes.

Now, should Twitter work on improving their algorithms to compensate for basic physics?  Yes.  But to blame it on "racism" shows just how divorced from reality some people are.


But you used reason and logic to point that out, as opposed to reductively blaming it on racism, meaning you are this threads 'Secret Hitler'.
Every thread has to have one, so that people who want to appear non-racist have someone to call racist to show everyone how non-racist they are, and you just nominated yourself.
You damned racist, you.
 
2020-09-21 11:10:36 PM  
Christ! When will the whining end? Did the white kid next to you get a bigger cookie? That must be a farkin conspiracy too! Maybe let some shiat go if you want to be taken seriously.
 
2020-09-21 11:13:45 PM  

bthom37: Pichu0102: This thread is going to be a real shiat show. Please make sure your dogs are wearing protective ear equipment before continuing down this thread.

Certainly you wouldn't suggest that the main page is a stronghold of the Kentucky Klan, filled with subliterate mouthbreathers?

Because that would be accurate.


*Throws flag*
Offsides. Othering before the play. 5 yard penalty.

You have to wait until the threads Secret Hitler is nominated -before- engaging in excessive virtue signalling.
 
2020-09-21 11:14:34 PM  
Uh, no.  Some guy, a presumably un(der)employed self described "cryptography and infrastructure engineer" (a made up title) found a way to game the recommendation engine with photoshop.  Actually competent ML people are very aware of the risk of bias when they use historical data sets for training.  This guy is a clown looking for his 15 minutes of fame.
 
2020-09-21 11:36:34 PM  

AmbassadorBooze: Shutdown Twitter.  Problem solved.


No way. Twitter's awesome


Mark Hamill likes TWO of my tweets today. We're friends now. And it's all thanks to twitter.
 
2020-09-21 11:47:15 PM  

tasteme: The Pope of Manwich Village: tasteme: Michael Jackson said it doesn't matter. Obviously Twitter is bias against white women

Yeah, yeah, yeah.

You're singing it wrong. It's yeah, he, he, aow!
Kids these days, jeez


It's black, it's white
It's tough for you to get by (yeah, yeah, yeah)
 
2020-09-21 11:51:09 PM  

alechemist: Saw this earlier on Twitter with lenny and Carl used as an example  .


It's even doing it with dogs.
 
2020-09-22 12:11:46 AM  

The Pope of Manwich Village: I'd pay extra for a freckle filter.


A good set of freckles is an awesome thing. I love speckled people.
 
2020-09-22 1:04:53 AM  

Kalyco Jack: AmbassadorBooze: Shutdown Twitter.  Problem solved.

I've been saying this since they started. It's really telling when people think they can have intelligent discourse in a medium that limits you to 140 characters.


280. Also, just guessing with no data to back it up.... If Twitter users are majority white, then it would make sense that more white people photos show up.

Regardless, fork Twitter. I would love it if it was burned down.
 
2020-09-22 2:03:22 AM  

Dryad: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

Programmers for years have been saying two very basic things, both of which get tossed out in the crush to call them all racists;
-
1) Sensors don't take pictures of objects. They record light. Specifically the number of photons bouncing off of objects. Objects that bounce fewer photons leave the sensors with much smaller sample sizes to try to interpret, and much smaller deltas to compare. Which leads to;
-
2) Computers don't see people. They mainly use tools like contrast detection to try to identify general shape.
This is harder to do if there reduced contrast between lighted and shadowed parts. This is as true of less reflective objects as it is less reflective people.
-
/In short photons are racist and we should boycott the universe until it comes up with more racially sensitive energy emissions


And the people biatching about this conveniently skipped over the bit where they said the problem is the contrast between an individual's face and smile.
 
2020-09-22 3:56:04 AM  

chawco: AmbassadorBooze: Shutdown Twitter.  Problem solved.

No way. Twitter's awesome


Mark Hamill likes TWO of my tweets today. We're friends now. And it's all thanks to twitter.


I hope you factored him liking you into your Xmas shopping.
 
2020-09-22 5:01:57 AM  

Dryad: nobody11155: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

I strongly suspect it was nothing of the sort.  It is kind of a PITA to process a photograph of dark ANYTHING.  Go take a picture of a black lab next to a yellow lab and see which one is easier to process to detect all of the detail.  This is just simple reality.

That the results had a color bias has nothing to do with "racism."  They are going to have the same problem with someone blond and blue eyed tanned to a burnt crisp.  Is is just a side effect of the physics of the visible light spectrum and how we tune our images to the peak reception of our eyes.

Now, should Twitter work on improving their algorithms to compensate for basic physics?  Yes.  But to blame it on "racism" shows just how divorced from reality some people are.

But you used reason and logic to point that out, as opposed to reductively blaming it on racism, meaning you are this threads 'Secret Hitler'.
Every thread has to have one, so that people who want to appear non-racist have someone to call racist to show everyone how non-racist they are, and you just nominated yourself.
You damned racist, you.


You are the Godwin AICMFP
 
2020-09-22 5:36:32 AM  
nobody11155:
I strongly suspect it was nothing of the sort.  It is kind of a PITA to process a photograph of dark ANYTHING.  Go take a picture of a black lab next to a yellow lab and see which one is easier to process to detect all of the detail.  This is just simple reality.

Ya, we have a mostly black cat. All pictures of her pretty much suck. Try to get a cute picture of her in your lap and it's just a black blob, not cute like a furry blob.
 
2020-09-22 5:44:08 AM  
Well, theres more whites and theres twice as many boobs as women, almost.
 
2020-09-22 9:02:38 AM  

nobody11155: Chromium_One: Hunh.  When I first saw this story something more than a decade back, the result was blamed on the majority of involved researchers being white and algorithms being trained against datasets that included mostly white people.  It was at least vaguely plausible.  Unfortunate that in this iteration all they're saying is "But it's clear that we've got more analysis to do."

I strongly suspect it was nothing of the sort.  It is kind of a PITA to process a photograph of dark ANYTHING.  Go take a picture of a black lab next to a yellow lab and see which one is easier to process to detect all of the detail.  This is just simple reality.

That the results had a color bias has nothing to do with "racism."  They are going to have the same problem with someone blond and blue eyed tanned to a burnt crisp.  Is is just a side effect of the physics of the visible light spectrum and how we tune our images to the peak reception of our eyes.

Now, should Twitter work on improving their algorithms to compensate for basic physics?  Yes.  But to blame it on "racism" shows just how divorced from reality some people are.


So you completely agree with the point, but are arguing because of the voices in your head. Got it.
 
2020-09-22 9:03:18 AM  
Proof:
Fark user imageView Full Size
 
2020-09-22 9:49:55 AM  
There must be something wrong with the way my PC displays tweets. All I see is people posting the same image twice and then saying "Look at the racism inherent in the system!"
 
2020-09-22 10:51:56 AM  

dready zim: There must be something wrong with the way my PC displays tweets. All I see is people posting the same image twice and then saying "Look at the racism inherent in the system!"


You got it. That is the entire point.
Actually addressing and rectifying the racism inherent in the system is HARD. It's like, work and stuff.
Tweeting pic of literally anything and saying its racist is much easier and doesn't make wypipo people late for goat yoga.
 
2020-09-22 2:12:42 PM  

thurstonxhowell: Kalyco Jack: AmbassadorBooze: Shutdown Twitter.  Problem solved.

I've been saying this since they started. It's really telling when people think they can have intelligent discourse in a medium that limits you to 140 characters.

... and, yet, science Twitter has been enormously useful to me professionally and intellectually. I don't think I follow anyone, except MC Hammer, without a post-graduate education.

It turns out you can do quite a lot with short messages. Also, we do this new thing now on the internet where you click the text and it takes you somewhere else. You should check it out.


OK, limit twitter to only posters with post grad education.  Problem solved, with the new constraint.
 
2020-09-22 2:49:42 PM  

dready zim: There must be something wrong with the way my PC displays tweets. All I see is people posting the same image twice and then saying "Look at the racism inherent in the system!"


It's not the just same image twice. It's kinda four images. Picture A above Picture B, then Pic B above Pic A. Twitter then chooses to show the same pic twice.
 
Displayed 44 of 44 comments

View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking





On Twitter



  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.