Facial recognition software regularly misgenders trans people [shock]

Ok Pollydron, now you’re going to have to tell us more about your job because that sounds fascinating!!

We are all human facial recognition computers, and we know a man when we see one.

What about people like Pips Bunce and the police officer who has warrants under the names of Abi and Colin on different days? How is the machine supposed to differentiate between their alter egos?

OMG. Someone call a waaahmbulance.

Bollocks article. Once upon a time, I worked in a woo woo job and was trained in facial recognition and used FR software daily.

FRS is designed to be ‘gender’ blind. It’s purpose is to identify an individual, not be obfuscated by things like makeup, hairstyles etc. Even changes from HRT and minor facial surgery barely make a difference.

I saw many examples, both in training and irl, where it successfully identified an individual who’d changed ‘gender’. Also examples where it got it wrong because it matched a natal female and a natal male who were definitely not the same individual but had very, very similar facial characteristics.

Fun fact for any tin-foil-hatters: your ears are actually more important in this type of process than the ‘gender’ written on your documents.

I think it may be an urban myth, but it demonstrates a truth about AI systems.

Errol you really piqued my interest. On training AI to recognise tanks:

In the early days of the perceptron the army decided to train an artificial neural network to recognize tanks partly hidden behind trees in the woods. They took a number of pictures of a woods without tanks, and then pictures of the same woods with tanks clearly sticking out from behind trees. They then trained a net to discriminate the two classes of pictures. The results were impressive, and the army was even more impressed when it turned out that the net could generalize its knowledge to pictures from each set that had not been used in training the net. Just to make sure that the net had indeed learned to recognize partially hidden tanks, however, the researchers took some more pictures in the same woods and showed them to the trained net. They were shocked and depressed to find that with the new pictures the net totally failed to discriminate between pictures of trees with partially concealed tanks behind them and just plain trees. The mystery was finally solved when someone noticed that the training pictures of the woods without tanks were taken on a cloudy day, whereas those with tanks were taken on a sunny day. The net had learned to recognize and generalize the difference between a woods with and without shadows!

www.jefftk.com/p/detecting-tanks
It looks like it might be true, but v old (60s)

And we have just solved the conundrum of why Cyberdyne Systems [TM], on reaching sentience, decided to kill all the humans.

The point with AIs is that you don't deliberately program to detect makeup as a proxy for gender - but if the photos are tagged with 'gender' eg masculine, feminine or neither, then there's a good chance that fairly accurate assignments could be made which would actually be relying on such gender stereotypical cues.

If only more women were more gender confirmstive. You could program them to detect gender using fake eyelashes and lipstick.

Computer says no.

Well, who knows what a neural network is going on? (there was the infamous, maybe mythical but plausible, case of an AI trained to detect tanks which actually detected cloudy weather really well).

If there were sex and 'gender' attributes, I'd think there was a fair chance of AIs distinguishing eg people with prominent browbones wearing makeup, or people with smaller facial features and stubble.

'Nonbinaries' ... who knows. Hairstyle and poutiness?

Puts me in mind of the Cultural Revolution in China, where machines were said to operate only by "Mao Zedong Thought" and not by mechanisms.

Erm, how would this software ever be able to go on 'gender' rather than sex? Bonkers.

Machines aren’t value neutral, they act as they’re programmed.

Actually, a lot of AI depends on neural networks etc, not deterministic programming.

I assume they'll generally use a large training set of pictures each of which has a 'sex' or 'gender' attribute. (This is the level at which things may not always be 'value neutral', I suspect - what attributes are being tested?). A possibility would be to include both (though probably harder if the options aren't binary?) - and then see how the results differed.

I'd take a guess that if the training data is representative and has appropriate attributes, AI might be able to accurately detect some transgender people using the same cues humans do unconsciously. whether that's the outcome those people would want, I don't know...

All. The. LOLs!

Keyes doesn’t see a need for any kind of AGR at all.

“Technologies need to be contextual and need-driven,” they said. “What are the values of the people who use the space that you're deploying a technology in? Do the people in that space actually need it? If we're not discussing gender at all, or race at all...it doesn't necessarily lead to a better world.”

But you can't solve the racist AI problem without actually discussing race. If you pretend race doesn't exist, you'll get racist inputs and racist outputs because we're in a racist society.

The "genderfucky nightmare" wants sex/gender to be ignored. They want it to not even be a variable. Very disingenuous of them to pretend that the "gender" problem and the "race" problem are analagous when they think the gender problem can be solved by ignoring it.

Facial recognition should be informed that if you wear pink, you're a feeeemale :P

Homestar

All the AI needs is labelled data to train with.

Good luck with getting the non-binaries to agree on a set of labels

Keyes found that researchers followed a binary model of gender more than 90 percent of the time,

I'm now laughing very hard at the idea of trying to build a facial recognition model that can reliably identify nonbinaries.

Seriously though, this seems laughable when it's just facial recognition software but how long before they start attacking researchers for including a "binary model of gender (sex)" in health research?

Whatisthisfuckery eek...you've asked the question, so that has invoked The Trans. I guess a bit like saying 'Bloody Mary' three times in the mirror.

You'll need some blue hair dye.

FunnyTinge hm, well I thought I was female, because of vag and giving birth and that, but I do like some boys’ things now you mention it, and my hair is very short. Now I’m confused. If I google ‘am I transgender?’ Do you think I might get some useful advice?

And yet... AI software has been shown to be sexist and racist. Does that have "immediate consequences"?

tech.co/news/sexist-ai-doomed-reflect-worst-2018-10

Whatisthisfuckery are you quite sure that you weren't assigned female at birth by mistake?