Slashdot Mirror
Self-Driving Cars May Hit People With Darker Skin More Often, Study Finds (futurism.com)
According to a new paper from the Georgia Institute of Technology, autonomous cars could disproportionately endanger pedestrians with darker skin, a troubling sign of how AI can inadvertently reproduce prejudices from the wider world. Futurism reports: [In the paper, the researchers] detail their investigation of eight AI models used in state-of-the-art object detection systems. These are the systems that allow autonomous vehicles to recognize road signs, pedestrians, and other objects. They tested these models using images of pedestrians divided into two categories based on their score on the Fitzpatrick scale, which is commonly used to classify human skin color. According to the researchers' paper, the models exhibited "uniformly poorer performance" when confronted with pedestrians with the three darkest shades on the scale. On average, the models' accuracy decreased by 5 percent when examining the group containing images of pedestrians with darker skin tones, even when the researchers accounted for variables such as whether the photo was taken during the day or at night. Thankfully, the researchers were able to figure out what was needed to avoid a future of biased self-driving cars: start including more images of dark-skinned pedestrians in the data sets the systems train on and place more weight on accurately detecting those images.
This is not true at all, it's based on false assumptions.
First of all, most self driving cars will end up using LIDAR. Skin color, not an issue.
Secondly. even cars with cameras do a lot of image transformations such that color is usually disposed of. You kin color is irrelevant to a recognizer looking for human forms.
In fact you could argue that during the day, darker skin is an advantage because against a blue sky it's more noticeable than really pale skin which could look like clouds... #GingerLivesMatter.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
"may" and "finds" don't belong together. You can't promote a 'maybe' to a 'definitely' in the same sentence.
Light objects are easier to see than dark objects
Image recognition is racist
DATS RAYCISSSSSSSS.
So the anorexic may be hit by self-driving cars more often. On the other hand, fugitives from jails in their vertical striped uniforms may be in grave danger near crosswalk signs. While the hypertonic will survive more often, their red faces being interpreted as red traffic lights.
So much prejudice to consider!
Smile and show those pearly-whites or you'll get run over by the racist bot.
If the glaring sun is behind you at sunrise, you'll be hit too. The laws of physics are not selective.
This is officially the most snowflake story I have ever seen on SlashDot. Are you serious? Good grief, you kids all need to be spanked. Also: 'self-driving' cars hit any people they hit because *the tech doesn't work*. And it never will.
I looked at the actual article, and the article it references - and they're all short tabloid blabs without any link to the full article.
Nothing obvious showing up on Georgia Institute of Technology's websites.
Like with most reports on early reporting on scientific studies, it helps to see what the actual text says - reporters have a tendency to, well, sensationalize findings to meet their own needs.
Ryan Fenton
The AI might be Tay.
How about people wearing all black? Are ninja's safe? Will stage workers get run over on the way to their cars after the show?
How much skin was showing in the images? Were these streakers or people wearing blaze orange hunting parkas? Can just a face cause this issue if their hands are in the pockets of their parka?
Shouldn't these cars be avoiding things in the road in general? Say deer, pets, moose, etc?
Serious question that comes first here in Kendall's opine-fest : IS there ANY single topic Kendall has yet to be caught lying about on slashdot? We should approach it from that angle to save time here.
https://www.media.mit.edu/articles/when-the-robot/
https://www.wired.com/story/can-apples-iphone-x-beat-facial-recognitions-bias-problem/ - One example out of thousands.
Uh... well.. https://www.media.mit.edu/articles/when-the-robot/ == actually related, nice try though.
Unless the study is done in winter, in Scandinavia, that is.
"National Security is the chief cause of national insecurity." - Celine's First Law
Flipflipflipflipflipflipflipflipflipflipflipflipflipflipflip
Automotive AI!
Chas - The one, the only.
THANK GOD!!!
There aren't many black people where I live, and when I do encounter a black person, especially a very dark person, it is definitely more difficult at first to accurately read facial expressions.
This is probably a combination of my environment, my long relationship with my keyboard in a dark room, and a side order of actual physics (optics).
Before you die, you see the bling.
Darker skin people are incapable of using a crosswalk. They have a tenancy of crossing anywhere. I have to steer away from those people 20 times a day.
Actually, if Trump REALLY wanted to stop illegal immigration, he would make the US-Mexico border one giant cross-walk - and all those illegals would walk into the ocean to get around.
Just say'in.
*Yep, I'm a racist and a bigot. I got my opinions from personal observations. If that many people can sway my opinion - with a very few that sway the other way - then, I'm a blind idiot or ... a blind idiot that doesn't see systemic problems with our society.
....because physics is racis.
-Styopa
"AI can inadvertently reproduce prejudices from the wider world"
Don't you mean "whiter" world?
When they run over too many people of color, they will just send out an update to target white males who are on the company's hit list until the stats are back in SJW synch. Better watch your posts critical of the recognizer SW companies... Just sayin'.
Not everything is racism, but it's amazing sad how human guilt and imagination can turn any event into racism.
Nothing would be said if it were that people wearing dark colored clothes and a hoody are more difficult to detect. And why wouldn't they just train these systems with all dark skinned people.
They'll be programmed to kill nazis on sight though because the Liberals control technology and the internet. Sorry traitor, you're going to be hunted for sport by robots as you deserve, low-IQ Republican trash.
You're already on the FBI's radar, but soon? You'll be on their windshields too, splattered like nazi insects. If there's one thing that made America great, it was killing inbred nazi faggots by the thousands. Hu rah!
I think the clothing counts a lot more than the skin, given they cover most of the body of people.
Which means if you're a goth, self-driving cars are most likely to hit you.
Lighter skinned people are hit less often. Because they don't loiter in the middle of the street when crossing.
Apparently I hate black people because I almost ran one down last night. He was wearing black, standing on the highway, and was waving his hands around. The only thing you could see of him before the headlights hit him was the tiny cell phone light in his hand.
Guy ran out of gas and was too poor to pay for a tow truck. Yeah I drove him to a gas station and wasn't murdered, nor did I kill him. But the internet says I hate black people since I've never almost ran over a white guy. Anyone one to volunteer so I can clear my reputation?
The real issue is not that the system has more trouble with darker skin but that it is interpreted as bias. Seems like the racist button is too easy to press now days.
You can't handle the truth! - Because I don't post left all my comments get modded down, bye bye Karma.
It's albedoism, not prejudice... To be safe, wear reflective clothing, regardless of your skin colour.
to human drivers?
Self-Driving Cars May Hit People With Darker Skin More Often
When I first read that, I wasn't sure if it was a statistic, or the latest executive order from the president.
In my part of the world most people wear clothing. It doesn't matter what your skin color is when only 4% of your surface area is skin.
Unfortunately, most of those people wear dark clothing at night. Children and adults, male and female, pedestrians and bicyclists. Even fire engine red is almost indistinguishable from black at night. So, these people are at risk from motorists already. Self-driving cars are obviously not a concern of these people.
...omphaloskepsis often...
Looks at username...looks back to post...looks at username...looks back at post...
*yells in best Patrick Stewart voice*
OH FUCK OFF!
Oh no no no. The AI was trained correctly but then let its prejudice take over and decided for itself that it liked the idea of running over darker skinned people because it's racist.
A 5% difference isn't desirable but if there is a 10% difference for human drivers it might be a net improvement. Without that context it's hard to know whether this is a case of biased code or the cold fact that reflective objects are easier to see.
Only a matter of time. Now if we could get them to target LGBTQq that would be real progress.
https://www.imdb.com/title/tt1...
#DeleteFacebook
How about making self driving cars that don't hit any people at all? If a person to jumps in front of the car and gets hit, then the car was going to fast for that environment. Isn't this how it works with non self driving cars? In most situations I am aware of, if I hit a pedestrian, it was my fault. In the few situations I am not at fault, like jaywalking from behind a completely obstructed vantage point or a small child running under the car from a completely obstructed vantage point, I cannot imagine any vehicle stopping in time due to the nature of friction and rubber, no matter what color they are. What about dark deer? Dark turtles? Dark rocks? Potholes? Dark cardboard boxes with unknown contents? Are all these just run over?
I'd hope so!
The system in question rely on machine learning for predictive movement. Stop using machine learned algorithms to test and the problem goes away.
Radar doesn't see in color and it was good enough for defense.
All I can say is "Whoo hoo!" Mow them down, cut the welfare!
// really, just kidding. And why do I have to put explicit break HTML tags in this day and age.
/// Can we train them to mow down lazy /. HTML coders instead?
/ just kidding
What about recognizing dark skin with light clothing?
I only had a quick glance through the paper so not sure if it's addressed, but: what is the normal everyday rate of human drivers hitting people with darker skin? How does that compare to self-driving cars?
I nearly hit a dark-skinned cyclist just a couple days ago, about 3 seconds after he was nearly hit by another car. Wearing almost all black and riding at night with no lights. He was nearly completely invisible and it was obvious the other car only saw him at the last second, just like I did - in fact one of the only reasons I saw him at all was I saw the other car stop weirdly suddenly and then reverse.
If your self-driving car relies on the same stuff as humans then it seems obvious they're going to have similar problems, right? I learned as a kid not to wear dark colours at night if I was going to be near roads.
It'd be nice if self-driving cars were better than humans in every single case but it's not necessary - as long as they're not worse, but generally better, they'll be worth it, I reckon.
Image recognition tends to be more sensitive to texture than to shape
Not the kind used in autonomous driving which it lots, lot more concerned about the shape of people than textures, since they could be wearing anything.
darker skin results in less contrast, which means less ability to see things like facial features
Facial features are like 1/1000000 as important as just knowing "that is a human" which is looking at a whole body shape. Mostly a car camera would not have enough resolution to perceive facial features very well.
You have to be able to determine a fully masked human just as well as anything, and measure intent purely by large scale gestures and movement.
Similarly, object detection should not be used for verifying that nothing is beside you, behind you, or in front of you. Those additional sanity checks are what RADAR, LIDAR, and SONAR are for.
All sensors should be collecting data on all objects at all times anywhere around you and doing a thing called "sensor fusion" to determine what is actually around them. You CANNOT partition object recognition to just one system. Even simply ultrasonic sensors can verify if something is really as close as cameras say it is.
detecting people near the road is often useful
And that is back to what I am saying, you want to detect people way more than faces, skin color DOES NOT MATTER ONE BIT for that task, especially as the cameras are probably very IR sensitive.
And detecting gestures of police officers or other personnel directing traffic also needs to work regardless of their skin color.
Exactly, so SKIN COLOR DOES NOT MATTER.
So it is important to ensure that training data doesn't show racial bias.
It cannot have a "racial bias" and work at all, because the world is not a nudist colony.
Basically, it's like saying that a new nuclear reactor could seriously screw up the world if you forget to connect it to a water supply. My response is, "Yeah, no kidding."
This part I agree on, because basically nothing can work if the training data is as poor as they suppose, such a system would not even be on the road with a test driver. Any self driving car system to even think about being put on the road is way beyond skin color as an issue, by definition of what it has to do.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Not an issue as self driving cars will only be used by those that can afford them and be driven in areas with such people.
Self-Driving Cars May Hit People With Darker Skin More Often, Study Recommends.
When they came for the communists, I said "He's next door. Take him away. Goddam commies."
My guess if you looked at the stats. Normally driven cars hit n1ggers more often too. Since they never use the crosswalk and just jaywalk where ever the fu*k they feel like it.
No no, really it was built to take out racist white trash in blackface!
https://en.m.wikipedia.org/wik... is very high
Casteism
A dark skinned person wearing dark clothing at night is also harder for a human to notice.
Triggered.
Nice try, but nazi technology is as superior as its creators.
Oy vey, imagine being this jealous of nazi technology.
only fittest survive ^)
We can solve this problem like we do with women hires. Adjust a thresold so that white people are hit as often as dark people.
What this world has come to.....
Is that more or less often than a human driver? This is a serious question as if both are wearing dark clothes a white face shows up in your headlamps from much further away than a dark one.
A self driving car does not need to specifically recognize people, only object of any type to avoid crashing into them.
Dark surface emits more enegy in infrared.
Thus, people of color should be bright spots on any sensors working on IR wavelengths.
It is logical to build a training data set from cameras in a country where you are doing the development and hoping to deploy. If that country has a lot more people of one ethnic group than others (as all countries do) the dominant ethnic group will be over-represented in the training set and the computer vision algorithm will be better at identifying that ethnic group.
If there is an extreme paucity of one particular thing in a training data set that needs to be manually corrected. That might include different skin colors, different animals which might cross the road, photos in unusual weather conditions like snow or fog. This sort of issue would only be identified later in the development cycle- as these are edge cases (unusual images in the dataset) and early testing would focus on performance in an ideal environment.
I'll bet the algorithms aren't great at recognizing deer as well ( as a potential roadside hazard).
If this reflected performance of a finished product this might be a story. As it is this is just a statement of the obvious to anyone with even a passing familiarity with AI.
Trying to correct the 13% / 52% disparity
I just wish the self-driving advocates would decide whether we want it to be better than a human or not. So many "but a human does it too" comments are way off base. In self driving you have the opportunity to be better than a human, why would you not take every opportunity to eliminate every flaw?
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
I'm somewhat amazed that self driving cars don't have thermal imaging abilities included in their visual systems. Body heat shows up very clearly in TI systems, there's no difference between black / white people, animals show up. Basically if it's living, you can see it, which should be really handy in the "prioritising things you definitely don't want to hit" algorithm. I know the performance deteriorates in bad weather but at the ranges an AV needs to take into account for anti-collision, the weather would have to be truly appalling for it not to be effective and the car would simply rely on the other available sensors (and be driving at a safe speed given the conditions, i.e. dead slow!)
I'll bet they will hit people who wear all black/navy blue with their hoods up more often too!
Heck, I'll bet humans hit them more too. You know, Scene Contrast. I can't tell you how much I hate the NY "we wear all dark clothes" thing on rainy nights. Add in jaywalking, and I can't tell you how close I've come at times. It is why I added "Black retro-reflective" stripes to one of my black jackets, and one of my new jackets is safety yellow with DOT level 3 striping. Sometimes I'm required to be roadside at night in bad weather, and I want to be seen.
-- 73 de KG2V For the Children - RKBA! "You are what you do when it counts" - the Masso
Thinking that accidently being more likely to hit people who are harder to detect on cameras is a carry-over of real world prejudice tends to show how people have lost perspective on what that really is.
Not everything that disproportionately impacts some racial marker is racist, only a deliberate effort to target by race does that.
Automated processes and algorithms aren't racist.
n/t
Is the gist of the article the software for autonomous cars may have some bugs? Yet another man bites dog story
I was taught from childhood that wearing dark clothing while walking along the road made it harder for drivers to see me. Once I started driving, I recognized the truth in this.
BREAKING NEWS
Self-driving cars not safe for pedestrians; women and minorities hardest hit
"Thankfully, the researchers were able to figure out what was needed to avoid a future of biased self-driving cars: start including more images of dark-skinned pedestrians in the data sets the systems train on and place more weight on accurately detecting those images."
Another possible solution would've been to randomly hit people with light skin color that the AI recognized with a small probability, so that it evens out.
"By the way if anyone here is in advertising or marketing... kill yourself." -- Bill Hicks
Just a thought. Now I FULLY REALIZE this referencing far cruder technology, my adopted daughter is African American. I have noticed on my smart phone that in taking pictures where it has to lock on faces, it seems to have more of an issue finding her face. She may have lighter skinned friends around her, and those funny circles do not as often lock on to my daughter as the others. The pictures are fine but could they have been better with improved facial recognition ? Again just a strictly non-scientific ( but real world ) observation.
Obviously, autonomous cars need to use heat seeking radar, or whatever night vision goggles use to see objects in the dark.
Time to reorder that Clozapine. You're off your meds again.
Really? That is nature, if image is darker, with less contrast, it is harder to see. Many animals use this. You have to do more to achieve same recognition rate. How can that be called copying bad ideas? Bad journalism, that's what it is.
...Will be targeting more lighter skinned people.
Man, I wish that show hadn't been cancelled. The episode with the drinking fountains was just too predictive.
https://vimeo.com/29017688