Follow

Indigenous faces don't count as human faces to Apple?

Digi-genocide, Techno-colonialism.

Dystopian fiction: "In the future, yt people *might* be treated like indigenous and people of color are treated."

Yt person: "Akxually that's a particularity of biometric algorithms????? DUH!!!"
Except there are already AAN-trained apps that can digitally erase tattoos for detection, plus there are failsafe work-arounds for indigenous Maori, Kanaka Maoli people so we don't get locked out of our devices with a 'That's me' button. It's just West-centric yt-cis-male dominated tech industry and data sets don't care about indigenous people and participate in their erasure.
Yt person: "No, ur dumb. Facepainting savage."
Ok racist.

I mean this is literally what IndigiStar deals with: the demonization and 'unprofessional'-ization of indigenous tattoos, patterns, beauty norms, gender, architecture, dress, et cetera as part of settler plundering of our lands, resources and people. Apple is continuing cultural genocide by artificially 'intelligent' means.

Actual Artificial Intelligence requires Decolonized Intelligence. You can't win an AI arms-race when you operate under Western barbaristic arms-race paradigms. Otherwise you get mega-servers full of Synthetic Stupidity and Software Sociopathy. youtu.be/mrrGnUdIxBU

@silverspookgames the dillemma of the future;
Should one get face tattoos to evade facial recognition or not to not stand out for humans

@jonossaseuraava @silverspookgames guarantee you the police face recognition system is trained on mostly non-whites. All the oppression of the AI age with none of the utility or convenience.

@silverspookgames I always think about how face recognition services often literally cannot distinguish between Asian faces, either, or Black users in dimmer lighting. Like the racism is baked right into the programming of these things.

@voorhees @silverspookgames In fact, it is! Facial recognition typically relies on neural networks from what I've seen, and those are typically trained on datasets consisting mainly of white people

@SigmaOne @voorhees @silverspookgames reminds me of Google classifying black people as gorillas in their photo app.

People were saying "oh well it's just a limitation of the technology, it was just an accident" no! That's now how AI works! What happened was they didn't consider black people/indigenous people/etc worthy enough to put enough of them in the training data. That's not better than them intentionally doing it!

@SigmaOne @voorhees @silverspookgames inb4 "the technology is still immature/in beta, they'll get better at identifying people of colour in the future!"

That's not how that works either! You have a responsibility to be inclusive to the best of your ability from day one! If you're not doing that, not exaggerating, you're better off not making the product at all.

And it's not like a giant tech company didn't have the resources to make it inclusive.

@SigmaOne @voorhees @silverspookgames these aren't pretend arguments I'm making up either. People genuinely say those things and think they're valid excuses for this.

@silverspookgames
> can digitally erase tattoos for detection

Any links? Sources?

@silverspookgames also, why not use the tattoo as part of the facial data for more security? It's apart of the face and almost always there for life.

@AgreeableLandscape Right? Tattoo-inclusive face detection would actually give you much *higher* accuracy if the software was tweaked for the unique characteristics.

@AgreeableLandscape because first of all, algorithm needs to distinguish a face from the background. It does so by first recognizing elements like eyes, nose and mouth. Based on their position it draws a bounding box around a face, then crops the image based on this box, so that only face is in frame. Only then it analizes particular features of this face such as tattoos. However, heavily tattooed face(and some specifically crafted face paint petterns) prevents recognizing of face at step 1, so yeah

@silverspookgames

@crystal @silverspookgames The Chinese government has developed facial recognition AI that can identify anyone, regardless of skin colour, while they're wearing sunglasses, or, as tested in the pandemic, a mask. It's definitely possible. It's just that Western techbro companies, in all their technological prowess, can't figure it out.

@AgreeableLandscape I don't know much about China's facial recognition systems, but I doubt they are much different from the rest of the world, I couldn't find any evidence in the internet that they are somehow more effective in any regard. However for me it seems weird to complain that surveillance systems that belong to shady corporations and corrupt governments cannot track you. Isn't it a good thing?

@silverspookgames
@silverspookgames no, it is due to how recognition works. It analizes light and dark areas of an image for a face pattern. Tattoos add extra dark areas which messes up the algorithm and it can't recognize a face.

@silverspookgames first of, I never said anything remotely racist. It’s just facts. Read more about the topic.

FR systems detect faces based on specific patterns of light and darkness. What the HyperFace camouflage does is mimic those patterns of light and dark in a way which looks like a face to computer vision, but not to human eyes. The goal of HyperFace is to make your real face a needle in a haystack for FR systems, whilst being relatively inconspicuous (beyond just wearing a cool scarf) to the people around you.

https://www.wired.co.uk/article/avoid-facial-recognition-software

Quite presumptuous of you to call me racist because I explained how tech works.

@crystal @silverspookgames simplest answer: then it needs to be changed. Humans should not be the ones conforming to technology, technology needs to conform to humans.

Besides being more inclusive for the sake of that, inclusive design makes it more useful for everyone. If you design it such that it can recognise a face with a dark tattoo, then it will probably also recognise someone in low light with half their face in shadow, for example.

@crystal @silverspookgames see the curb-cut effect. Having a ramp for people in wheelchairs to get onto the sidewalk will also help parents with strollers, or construction workers carrying wheelbarrows.

This is the same. You design your AI to recognise facial modifications, those same mechanisms will not only help the people directly intended, but everyone else too.

@silverspookgames related: In the West, body modifications like tattoos and piercings make you less likely to get through job interviews and are generally seen as very unprofessional. This is because they are associated with, one, indigenous people and people of colour, and two, poor people.

In contrast, other body modifications like nose jobs, facelifts, fillers, botox, etc are done mostly by rich white people, so they are seen favourably.

@AgreeableLandscape @silverspookgames well and specifically it's how well you're performing white supremacy. People aren't getting surgery to make their noses wide and rounded, if they're going for upperclass approval they're all coincidentally trending towards an artificial aryan/nordic stereotype. (Massive colorism issues among colonized nonwhites as well, with lightening creams and stuff.)

@wilbr @silverspookgames I've joked about "Aryanization surgery" before with white celebrities getting facial modifications, but honestly, it's actually happening.

Sign in to participate in the conversation
Octodon

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!