Zoom, Twitter hit by allegations of racial bias in their visual algorithms
Reports emerged suggesting Twitter and Zoom have a racial bias in their visual algorithms. A user noted that Zoom appeared to be removing the head of people with a darker skin pigmentation when they use a virtual background. This is not the case with people having a lighter skin pigmentation. It was pointed out that Twitter cropped thumbnails to favor the face of a white person over a black one.
I think it's just that more lighter skins people use the services so the algorithms are better trained to recognize lighter faces. The same way it can be hard to tell Indians, asians, and even white people apart if you're not around them much.
Darker is harder to spot while lighter is easier to spot. Can USA chill with their racial obsession? It has to do with shit load of stuff, the software, the lighting, the camera, no one cares if Jimmy or Susan is yellow or brown or pink...
The term racist/racism has been thrown around so much that it no longer holds any real significance to anybody except the mentally weak. Just accept that everyone on the planet is a rabid racist and just move on already.
This kind of thing isn't intentional. It's certainly embarrassing for the company though.
Bias implies choice. The fact of the matter is that algorithms rarely work as intended; amusingly, if the programmer had intended for there to be bias it would likely be more even-handed. Machine learning needs a lot of information to be accurate, and the reality is that most of the program's "training" was using lighter skinned individuals. Depending on where this "saliency" algorithm was created, that could simply be what was available.
Man, so now the PC Culture is canceling The Cancel Culture. This is getting very confusing. Now algorithmic computer programs are racist.
Point out the racism towards white people on social media and then we'll start talking .
Oh yet another fake racist outrage.-zz-zzzzzzzzzzzzzzz
Kodak sued for making darker skin harder to see in dark backgrounds. Obviously racism.
Algorithms that were fed mainly bright colored faces tend to recognise bright colored faces better? Amazing.
The silver lining is that the ability of facial recognition software to be used against you may be more limited the darker your skin. With the rampant 4A violations everywhere, I would consider it an accidental blessing...
Looking at various experiments on Twitter shows that it seems to be more than just whether it can see black people or not. Things like facial expressions, text, and background have a strong effect on what Twitter decides to crop. It's not racism, its just coincidence honestly.
Are the computer algorithms less biased to seeing skin color that BLM?
While reading this article I was reminded of a Scottish woman asking Alexa to play "Something's Cooking in my Kitchen."
🤣🤣🤣🤣🤦♂️
I can't help it that I blind 12 babies every time I show my face. Of course AIs are going to see my face more easily. Everybody does.
Get a life.
This is not bias. Facial recognition systems are not able to identify dark skinned people as accurately as light skinned people because dark skinned people have less variance in the biometric structures of their faces, compared to light skinned people. To AI, dark skinned people look the same / very similar to one another. It's not surprising that it struggles to recognize their faces when cropping photos. It's a limitation of the tech, and a result of biology. Not bias.