Slashdot Mirror
Facial Recognition Is Accurate, if You're a White Guy (nytimes.com)
Facial recognition technology is improving by leaps and bounds. Some commercial software can now tell the gender of a person in a photograph. When the person in the photo is a white man, the software is right 99 percent of the time. But the darker the skin, the more errors arise -- up to nearly 35 percent for images of darker skinned women, the New York Times reported, citing a new study. From the report: These disparate results, calculated by Joy Buolamwini, a researcher at the M.I.T. Media Lab, show how some of the biases in the real world can seep into artificial intelligence, the computer systems that inform facial recognition. In modern artificial intelligence, data rules. A.I. software is only as smart as the data used to train it. If there are many more white men than black women in the system, it will be worse at identifying the black women. One widely used facial-recognition data set was estimated to be more than 75 percent male and more than 80 percent white, according to another research study.
So at least white men have that going for them. We should be grateful for it.
Software is RACIST!!! Fucking white men.
much to their dismay
Sounds like the algorithm is picking up on subtle color changes for its classification. With a wider dynamic range it may do better. Perhaps a transformation from true color to a false color would fix it
For catching criminals then?
of white male supremacist oppression. What else?
So, when the white guy struggles to differentiate black people with an "you all look the same to me" statement, it's automatically construed as racist and derogatory.
Has facial recognition validated that statement to hold more truth than previously assumed, or is this the fault of the programming behind it?
Of course no one will assemble a "diverse" set of images to train computers; white males are the only group that can be profiled and discriminated against without generating a flurry of enraged hashtags. So there we go. Use white males photos because nobody will be offended, and then blame white males because the AI trained on their images recognize them better; win-win.
lucm, indeed.
all just look the same.
This is obviously a ploy by the intersectionalists to further their grip on power by suppressing the white guy in a "socially acceptable" way! *dons tin foil hat* Oh I'd better not have done that, 'cuz MIT research says that *attracts* government mind control rays. Oh shit.
Anyway, no of course there is likely no conspiracy (though THEY sure want you to believe that!) but this is simply an artefact of too many white guys in the field (hoist on their own petard, eh). But someone's gotta say it, regardless.
Uh, what's with the swarming helos here all of a sudden? They're all bla... eh, AFRICAN AMERICAN helos too. Funny, that.
White people's faces reflecting more light is problematic.
Does this mean it will only identify me half the time?
Only online white boys are at all worried about "a flurry of enraged hashtags". If you can tear yourself away from pornhub for a moment look into facial recognition in China where they are not training their AI to catch MRA sex tourists.
This will all even out over time. There are a finite number of facial types by race.
Always knew that machines were bigots.
We has it.
Computers, which as we all know, are not subject to human error, decide to pay more attention to white males, because computers, unlike human beings, have no problem recognizing the importance of the white male, and thus we, human beings, rather than raising questions and doubting the all-mighty all-knowing machine, should just knuckle under and respect the white man like computers do.
So the government will be less likely to know where I am.
This reminds me of that ridiculous article (and accompanying video) saying that color film was biased towards white people. Around 3:30 in the video they have white and black people stand in front of a face-following camera and it doesn't work for the black people. Everyone acts like this is some sort of Harry Potter wizardry against the black man keeping him down when it's vastly simpler than that.
For progressively darker skin, progressively higher light on that skin is required to reveal its contours. The fundamental problem is that white and light-skinned brown people have their normal skin color shades in the midtones when a scene is properly exposed while darker-skinned brown and black people are closer to shadows. To expose properly for facial recognition of dark brown or black skin, you have to overexpose the midtones to bring up the shadows. Since people rarely take photos on purpose that are exposed for the shadows while blowing everything else out, it should be fairly obvious that facial recognition (and early ISO 32 color film and small-sensor cameras like webcams and phone cameras) will have a very hard time with dark skin. Sure, it could be a lack of data in some instances, but it's far more likely to be the fact that the skin absorbs more light and photographs are generally exposed too low to reveal enough detail for the machines to analyze.
If you think this is "racist" you're saying that the nature of light itself is racist. I don't feel like I should have to explain why that position is really stupid.
I prefer to think of myself as melanin and chromosomally challenged. Whose fault is it that I got a broken X? I blame Trump.
Looks like everyone else will be tracked and profiled as they move around in the real world. Only darker skinned black women will have a shred of privacy.
the camera sensor work with light and since a paler skin reflect more light than a darker one the signal to noise ratio increase and more features will be visible and hence the program can do more of it's magic.
Assuming you don't measure the whole scene and use the same exposure you could expose the image for longer (and get something more blurry due to movement) or raise the gain (and gain more noise) but the simple technological fact is that the darker skin will never be as easy to capture as the brighter one. Eventually both may still be good enough but that's the fact.
As for whatever the algorithms somehow are worse for darker people I don't know. But to conclude that from a poorer result on them isn't evidence by itself, especially since there's a totally believable reason and prediction for why this would be the case.
Similarly for women due to makeup their and hair work their appearance may change more plus also be harder to process both naturally and because they blend their skin to a smoother tone using make-up. Camouflage and disguise if you so want.
Put a brown bag on your face and see how discriminating the algorithm become .. Then it won't have a clue!!
Light is racist and sexist. Light needs to be fired from its job, and kicked out of college.
For recognizing people in a crowd, 99% is nearly worthless. That means if you are looking for one terrorist in a train station that services 300,000 a day your false positives are going to outnumber your actual hit 3000:1.
Addressing the actual issue, the move to universal facial recognition is just Silicon Valley engineerlets latest in a long line of stupid decisions stretching over years at this point (case in point: no, Tim Cook, slapping AR on cute emojis is not going to 'normalize' it for people prior to your sticking it everywhere anymore thsn snapchat, nor make your other apps or gimmickey tech any more useful down the line. Most of us are not that stupid or easily manipulated, and gimmicks get old). Get off the drugs and the ego, kids.
That said: Seriously? It isn't 'fool proof' for *anyone*. It was a stupid decision to integrate it into EVERYTHING. All you are accomplishing with your post is to confirm your status as a snowflake. There are legitimate reasons this has been a dumb move, a racist agenda is not one of them. Way to cloud an issue with your own personal butt hurt. Sorry, it may break your tender little heart, but this isn't about race, and it certainly isn't about *you*. Get over yourself.
As an addendum, this is not me or anyone else being a 'luddite' or 'arrogant'. You are legitimately a moronic person if this post represents your highest level of thought or scrutiny. Like it ir not, 99% of the time we really do know better than you do, so it has always been for grown adults and post-pubescents, and so ever shall it be.
Then there is no discrimination. Kinect infrared for example does a great job leveling the playground.
Until I was a little older I was utterly unable to distinguish between black people.
I'm not saying that as some kind of racist joke it is the truth, my adolescent mind just couldn't do it. Later on as I got older it became extremely easy to tell black people apart. Growing up though...well I feel for the AI, most movies with black actors in it really confused the hell out of me as a child since I couldn't tell them apart.
I had a similar problem with asians but it was too a much lesser degree and like with differentiating black people it wore off as I got older.
That was Google Photos IIRC.
That's why I went with the Iris option on my phone....
If you're a white guy, you have a lot of worry about, "they" can track you. If you're not, you're lucky, you can still blend into the creeps.
Another racist (Joy) concluding that the cause of something not considered ideal must be due to racist white men. Pure rubbish.
Computers know better, they do not come when the societal biases out of the box. You have to manually damage them algorithms to give wrong results.
Muh white privilege.
Someone needs to test whether humans, also, decline in speed or accuracy of facial recognition when dealing with darker shades of skin colour.
I know for certain that I have more trouble reading facial emotion from black people than white people. The naive response is that I live in a city that's 95% white. But I've been able to convince myself that this is the correct explanation. I simply feel like I have less visual data than I would otherwise at the same point in the cognitive process.
Suppose I lived in a troop deployment in Afghanistan, and 90% of the people around me wore camo all the time. Would I actually become better at recognizing camo than civilian gear? But this is, indeed, the converse implication of the naive hypothesis.
There are populations in Brazil that experience the entire range of skin tones on a daily basis. These populations could be tested for recognition rate/accuracy for lighter and darker test cases.
I highly suspect that darker skin tone has a detectable coefficient of identity camouflage, also in human cognition.
In our increasingly Orwellian society, I would be quite happy to have facial recognition technology be less effective on my skin tone (fair).
A closed mouth gathers no foot.
check your white privilege, video slaves!
Facial morphology refers to the various traits and features in a face. For example, the distance between the eyes, or the eye slant, or cheek gaunt or whatever.
'White' people have the broadest range of diversity, in part because aside from the skin color, there's a lot of differences. Certain Asians, like the Han Chinese, have some of the least diversity (google for iphone face recognition matching two Chinese co-workers).
If you pick 20 key features as your unique code, and each of those key features has 20-30 distinct possible values, you can rely on reasonable uniqueness, even when some of those values have inter-relationships. When the diversity goes down, and 10 out of the 20 are not unique, and when the range of values those have is between 3 and 5, well, you'll have a lot more trouble differentiating people.
In fact, a studies shows that among a given ethnic group, actual real life people perform facial recognition on only a few features, but those features are always those traits that show the most variation. When you apply that same algorithm to another ethnicity, it doesn't work so well. You get racist-seeming phrases like, "They all look alike to me," when really the issue is that your specialized detection algorithm was never meant to deal with their differences. ... and every group has this blindness. The one thing that's amusing is that because whites tend to have a large variety, they're the easiest to uniquely identify regardless of your personal/cultural/ethic technique. So, you can say things like "I can tell all you white people apart, you're racist for not being able to identify ME!" and think you're on the moral and ethical high road, when in fact, the situation is different from the other side.
Darker colors provide less contrast. Less contrast means features are more difficult to make out.
Combine that with the typically horrendous lighting video cams face and you have a situation where recognition fails.
That is a reasonable hypothesis, but we know cross-racial identification is hard, and training sets differ, so it is not the only reasonable hypothesis, and we would need to design a good study to verify it.
Rather than:
some of the biases in the real world can seep into artificial intelligence, the computer systems that inform facial recognition
Maybe the issue is lighting? Why does a simple thing such as AI to identify gender from a facial camera have to be an example of latent racism? As if programmers subtly, unconsciously, monkeyed with the algorithm to only work for white faces.
Ken
That's not the problem. The problem is the training set data. The algorithms weren't trained on many Gorilla photos, so they were extremely poor at recognizing them.
It doesn't work exactly this way, but consider:
1. you're given a photo a photo to classify
2. So you compare it against the things you've already classified
3. One of your rules is that categories that more commonly appear are more likely choices than categories that rarely appear.
4. You compare it against your match set, but it doesn't exactly match any of them
5. So you pick out certain feature:
a) It's sort of vaguely human shaped
b) The skin color is dark
c) What the most common picture category that matches that?
6. So you identify a gorilla as a black man, because pictures of gorillas are really unlikely.
As I said, it doesn't exactly work that way, but that has many features of what caused the problem. But an example of how it doesn't really work that way is the picture of a plastic turtle that gets identified as a rifle.
I think we've pushed this "anyone can grow up to be president" thing too far.
The geometry is found by distinguishing between recessed a car with as, which are dark, next to relatively high areas, which are brighter. When the entire face is dark, it's difficult to distinguished shadowed areas, and therefore geometry.
Invert the colors, look up in a different neural network and find matching patterns. This is not a technical problem.
If you have an issue with contrast ratios, improve your lighting or capture process.
n/t
Now even facial recognition wants to be racist, nothing is sacred on the land of the free anymore.
Before it was tacky but now it has a practical use. Stop the government from recognising us.
I guess the white/asian guys who make the software don't have black girlfriends? Only two ways to fix it. Date a machine learning researcher/coder or DIY.
Whites smell like spoiled milk to me.
Why are we calling this bias? White males have the most range of unique identifying characteristics:
* Beards
* Moustaches
* More tonal contrast
* Difference in eye and hair colour
Apple's FaceID uses infrared depth perception, where light contrast isn't an issue.
Apologies for the inflammatory title https://www.gizmodo.com.au/201...
They also went to the effort of testing it out on various ethnicities as well, so the AI didn't overly focus on areas that are different for one group but similar in another.
I'm not saying that Face ID is "racist" or anything like that. I'm just happy there's a technical solution that solves this problem.
It's turtles all the way down.
it may be hard for algorithms to distinguish detail on dark skin colored faces but recognizing Chinese people by AI is really marvelous state of the art that no white man can top .
Even computers think that all black people look alike!
niqqers dont have enough genetics to have major facial differences that why even for non computers it is hard for people to tell one niqqer from the other.
Niqqers work like all the other genetic inferior animals and sniff each others butts to identify each other.
Cunt from the meme "So what you're saying is all black people look the same?
Wasn't it the opposite scenario? Black men were identified as gorillas?
What is best in life? Hot water, good dentishtry and shoft lavatory paper.
I'm having trouble with the fact that it's so accurate at identifying men but not women. Gender politics aside, how does this work?
so, even computers have found that blacks and Asians all look the same.
Posting as anon for obvious reasons
You learn in art school that drawing (white people) is a critical matter of getting the eyes, nose and mouth right. You can literally draw ONLY eyes, lower nose and mouth and easily recognize the person being drawn.
The thing is these are NOT the parts of the face that are unique for Asians or Blacks. For Asians, it's the outline of the head that's more unique (i.e. carries more information that differentiates individuals by appearance).
So if you are using neutral nets, but especially convolutional NN, you can easily get this wrong by the wrong training sets and by the wrong pre-processing.
In some ways, more anonymity by failed recognition means whites are more watched and surveilled but these others are more prone to false positives and negative results of that.
. . . is all you'd need to know this. My wife is black, and Prime Photos tags dozens of random, unrelated black women as her. What's most frustrating is that you can't override or correct any of its mistakes.
It's not racist to point out that white males are the only group that can be objectified and labeled without spawning a twitter lynch mob. It may or may not be accurate, but it's not racist in any way or form.
People like you who associate anything and everything with racism are a social cancer that is going to ultimately desensitize people to actual racism. Feel free to maintain your current trajectory of virtue signalling and self-righteousness but fundamentally you're an impediment to social progress, nothing else.
lucm, indeed.