techno-racism? Legit or bulls**t? You - make - the - call wordforge!

Techno-racism: People of color's new enemy (msn.com)

discuss! Or fling poo, I'm down for anything.

Well, we know the Youtube algorithm shoves alt-right dog-whistle white-supremacist horseshit to the front unless you manually pound away with "stop showing me this channel" commands.
Try to find an unbiased neutral Star Wars/Star Trek review, you get a motherfucker with "the feminist agenda strikes again!!" in four clicks, and "Discovery is the great replacement" in six.
Garbage-in-garbage-out.

Unless you think that someone has invented it and that it's not a real thing, how is this even a question?

Only Klan rallies are racism.
Everything else is failure to pull your bootstraps.

It seems to be completely possible that facial recognition software could misidentify people based on photos with shadows and differing skin tones and hair colorations of darker skinned people. Play with the contrast on your TV and watch how things change differently for black and white faces. This is a computer system based on converting different colors to digital data from pictures that could vary based on how the sensors interpret colors as numbers.

You have to remember the computer does not see colors in processing. It only gets fed numbers that are interpreted by varying algorithms based on differing software that interprets the data differently based on who programmed it. Not to mention different scanners may give differing results based on the sensors that read the reflective light from photos or the camera sensors.

In a predominantly light skinned society the techs who make these things might simply be lazy and produce things that are better able to identify white people because it is easier, and screw the problems with darker complexions simply because it might require more effort. When success is probably determined by a process where mistakes are quantified, and the developers know that dark skinned people give more errors they would simply just use lighter people for their tests so they get less errors and they can sell their product. Why work to fix a problem when you can simply fix the test so as to avoid the problem.

I would label it something different than simple racism as it is probably far more associated with the american way of cut corners to lower expenses until they regulate the industry to fix laziness. Yay lazy white capitalists for making that the american way. I guess in a way we learned that from lazy ass white idiots who could not do it themselves so lets get brown people as slaves and pretend we are hard working people rather than wastes of white flesh who could not work a hard day if they were whipped into it, or as we call them @oldfella1962 whites.

The issues regarding facial recognition and PoC are well-documented and not at all made-up bullshit. Machine learning relies on a training dataset. Companies have been lazy and used facial images from the web, which are predominantly white and Western. A black Uber driver lost their licence because Uber use Microsoft's facial ID algorithm, which failed to correctly recognize them:

I've just spent the morning doing my practical exercises for my Open University module, which focus on Robotics. This involved programming a simulated robot to navigate a course and stop when it saw a red background, but not yellow. The "Reflected Light Intensity" input reports only the Red component of the RGB spectrum, so you can't use that as red is 255,0,0 and yellow is 255,255,0 (Red component is the first value in the trio). Full Reflected Light Intensity sounds better as it's the mean of the three values. But running the robot across different colours shows that some colours can have the same mean even if the RGB components differ (0,255,0 would be green, but has the same mean as 255,0,0 which is red, or 0,0,255 which is blue).

Measuring the colour components individually works, but setting it to "if background = 255,0,0 do X" only works if the background is perfectly red - in actual images, there's noise from the camera tech.

Now imagine applying this to detecting skin colour. And then, if you can get the computer to do it, telling it it needs to flip it's assumptions for a face that has a different skin colour.

This. I always get right wing alt-right racist Youtube content creator recommendations regardless of my preferences, and no matter how many times I tell it I don't want it, don't want to see it, and block some of those same creators, it always brings more.

If you think about it from a cosmetic POV you have products you use to offer up changes in your your mind interprets things like shade, size, depth, and all sorts of other things. Now you have a computer algorithm that has to analyze something in 3 dimensions to match up a face based upon the same 2 dimensional system as a picture. It is using the same things that can fool our own eyes to try and interpret a face based on numbers. How do you program the same gestalts our brain uses to construct shapes and recognize them into a computer based on colors that can vary based on make up and light factors?

You are going to have to develop some way of determining the skin tone of different races because certain skin features appear differently between races. Things like comodones, scars, and even the changing shadows under the eyes appear differently on different skintones. Browns, blacks, and blondes all have differing amounts and colors associated with them. You could spend years as a colorist trying to learn the different colors in hair to cancel out tones and shades and establish new ones. Then think about where your facial recognition color system would be if I altered my eyebrow color, changed my lipshade, or altered my facial hair like waxing my eyebrows or shaving. If it is based on color or shading I can fuck your recognition up by simply wearing a different concealer, blush, or eyeshadow.

I get ticked at every news search site. When I say no NY Post, faux news, or some other right wing lie place that means I never want to see that fucking shit again. But those sites pay them to put their content in front of me so they are always popping up anyway. I do not want white nationalist propaganda. I do not want @oldfella1962 racist lies pushed on me in my news. I cannot imagine how annoying it is as a black person to have faux news and tucker carlson be forced upon you even though you never want to see it. What is worse is I barely see anything from MSNBC or even a progressive news site put out there. I have never had to take a left wing site off my feed, and the only one that pops up regularly is Huffpoo. They are about as left as a log cabin republican.

That is pure racism to have the confederate traitor news pumped into your feed even when you do not want it. Why can't I keep things like The National Review which is pretty much the white nationalist Hitler Review out of my news when they are calling for lynching of black people, atheists, jews, GLBT people, and anyone who does not speak english. These people review things like what is the best tree to hang a black person from, or where is a good fence to leave a gay person to bleed out. But hey, every so often they have to test if I want to hear the good news about the third coming oft orange Jesus.

That is blatant racism, and not technical racism.

What kind of garbage are y'all watching on YouTube? Here's my "recommended."

Yeah, it's a real and very well documented phenomenon.

Do you have an opinion on it?

It’s significantly more nuanced than “racist data from racist history = racist tech” that the article implies and quotes. Some of it is Simpson’s paradox (albeit with far more curvy lines than usual), some (probably most) of it is a sort of laziness - not bothering to get a sufficiently diverse data set (this is a problem worldwide, not an artifact of a particularly racist history - Nigerian data sets often have the opposite problem, publicly available Chinese facial data sets are nearly all Han Chinese, though I imagine the ones they use for internal security are more complete. Apple’s facial data set is *probably* the best. And it’s *hard* so just calling it straight up laziness isn’t quite fair. You need to have balanced samples, not proportionally representative samples, and for each feature, not just skin colors.), and some of it is genuine inherent difficulty; darker skin is lower contrast against dark backgrounds (especially in the blue and green wavelengths). The last is not a thing that would have been trivially resolved if only Silicon Valley had been in Lagos, for instance.

Seriously? At 59 years old I could still work circles around you and barely break a sweat. As for having slaves do our work I agree that was a very bad idea. We (as a nation) should have picked our own cotton but hindsight is 20/20 as they say.

yes I do! If it can be done more accurately then a black based tech company should get right on it. Who knows, they may be doing just that as we speak. And god forbid their technology doesn't work as advertised for asian faces * then they will be in the hot-seat for racist technology.

* having lived in Korea for two years take any 100 Koreans of the same gender and approximate age and try to tell them apart without a lot of facetime (no pun intended) spent with them.
Just sayin'

(Meltdown Christian Bale)
Ooohhhh gooood fooorrr yooouuu!!!

This doesn't let you off the hook.

It's not so much the difficulty of getting the tech to work that's the problem as the fact that they will (although this is getting much better than it was) test it on a small subset of the population then declare that because it works for them it's ready for the market.

Video game reviews, and IT videos, mostly. The video game review scene is chock full of them, unfortunately.

1. Why should there have to be a black based tech company creating technology just for black people? Why wouldn't any credible tech company want to create products that work for everyone properly?

2. Asian people totally get screwed over by this technology as well. I've read about "smart" cameras that label pictures with Asian eyes as "bad" photos, because the algorithm thinks it's a white person thank happened to blink while being photographed.

the answer to question #1 is if black people were the most affected & upset by the technology they would be the SME's in what they are looking for in a technology that treats them fairly.
Your second comment is interesting, but not surprising that computers have trouble with this. But considering the various glitches in this technology I certainly hope that nobody is wrongfully charged or accused in any actual crimes just because of computer errors. Not saying I don't trust technology, but I'm not a fan of self-driving cars for example.

This happened to someone I know when she got her driver's license. She's Chinese-American, and the computer kept rejecting her photo, saying she had her eyes closed.

Anyway, this entire topic is a great example of how systemic racism -- or disproportionate negative impact based on race, for any hypersensitive white conservatives who may find themselves triggered by the "R" word -- does not require intent. It's legit, it's not bullshit, and it's all over the place.

pretty interesting choices! France really withdrew from NATO?

as for your friend's photo that is one good reason why sometimes it's best to do things "old school" rather than the latest/greatest technology. There should be some kind of manual over ride when we need actual human common sense to just take the fucking picture like they have been doing it for decades.

The article in the OP definitely attributes intent.

There's a few tech companies out there that I wouldn't be surprised if they did have intent. Like that one company that came out of nowhere and said that they were going to be rolling out their facial recognition system widely to anyone who wanted it. Those are some shady mofos. But even the inadvertent shit can be so bad that it might as well be deliberate.

When Facebook finally decided to open a facility to moderate content in Africa, they established it in only one country, staffed it with unscreened locals, and basically left them to their own devices. This wasn't exactly a smart idea. You had people with vast cultural differences monitoring reported posts from countries they knew nothing about or had a history of committing genocide against. They also didn't have any real rules or training in the rules. I believe that after it got some publicity they've done a few things to improve things, but still not nearly enough, and not before a bunch of people got killed.

It talks a fair amount about what happens absent intent.

There are certain areas where people do have to be racist. Cosmotology has a bunch of them because there is a difference in how certain chemicals work on different people's skin and hair. There is a reality that if you take a chemical hair straightener that would work on the natural haire of black people and put it on my hair for the same time my hair is gone. These are the same chemicals that are made to open up hair for color reception. Just washing, blowdrying, and styling the hair of a black person has to be done differently than a straight thin haired white person. It is just physics.

There is also physics involved in the measuring and converting of light into a digital format and processing it through number based programs. It is not terribly shocking you might need to develop different programs for different people. There are different filters used to touch up different skin colors. Those filters are different numerical conversions of the digital numbers that the display processes.

I am not a graphic manipulator, but it may be entirely possible that it will take different processing to determine matches for different races based on their physical differences and how computers process the elements of the face. Then it may be a matter of cost as to what your government wants to pay for while these things are being developed. i would imagine China and Japan may be far less interested in technology that can parse white and black faces over asian faces. In an asian country you might be able to eliminate the minorities through profiling and manual detective work because a black or white person might stand out for identification purposes. In america we really should be developing software that covers everyone because we have a diverse collection of people and even differing minorities and majorities that change via internal geography and a national system of law enforcement needs to be able to handle moving from Harlem, to chinatown, to chinbeard rural georgia.

I love that you have something from Down Periscope, underrated movie

Answer

1. There should be an expertise in ai in identifying different individuals in different races because there are different physical characteristics of those races that vary among the people who have them. We do have differing racial industries in things like cosmotology because there are legitimate physical differences in the different racial groupings. Not all asian eyes are the same, but they are different than other races eyes. The dark shades of asian hair include violets and greens that are different from black hair on black people which contain more reds and dark oranges. The textures and porosity of our skin and hair are different. Bone structure, fat distribution, and even changes in skin color are physically different for the different races. Black women get ashy dry skin, but white skin does not get ashy.

2. asians are a huge chunk of the world population even though they are a minority in the US. Asians can see differences in people from korea, japan, china, and phillipinos. Just like there are Irish, italian, greek, german, and english faces. There are black people from africa, the middle east, and carribean islands.

Facial recognition needs to be able to differentiate before we can use it for definitive ID purposes. I could see it perhaps used in probable cause warrants where you might be establishing a proper ID based on a probable match, but as a definitive ID they most certainly should not be used yet.