@laser That's like the intersection of the worst of
1) cishet bullshit
2) late capitalist marketing
3) algorithmic data analysis
@FuchsiaShock GDPR? I barely know her!
@laser @FuchsiaShock this is not a GDPR violation. The fact that you're suggesting as much suggests that you don't understand GDPR at all.
@FuchsiaShock @laser 4) Corporate Memphis
@papirbanan69@fribygda.no@laser I have no idea, but I'm pretty sure ACAB includes them.
@laser I'd say a sign to get the hell off of FB. The umpteenth one, I might add...
I trained an artificial neural network on thousands of images to try and determine what gender someone was based on what they looked like. Every time I tested it out though it just returned "It's not possible to determine someone's gender based on what they look like alone".
@laser I'm almost afraid to ask what the inputs and outputs of their API are...
@laser Genderize you data. Lol.
Integers are male, but real numbers are female. The strings are male if they end in a dot or start with a capital letter, otherwise female. Structs may be male or female depending on the contents of the structs. Vectors are all female even if they are integer components.
Θ @kescher@catcatnya.com@laser lots of dudes in Italy named "Andrea" will be very happy about this one
@laser with our hot new product you can cisform that problem you didn't have into data you could have just asked for
@laser Straights at it again as a Service
Θ @kescher@catcatnya.com@pixellight @laser my gender is NAME_NOT_FOUND
@pixellight wow, I did not know that was a thing
@pixellight @laser `Gender\Gender::ERROR_IN_NAME` and `Gender\Gender::IS_A_COUPLE`, the 2 genders
@pixellight @laser oh wow that is so strange why is Ireland a gender?!
@pixellight @laser oh wow its actually real. What were they thinking?
@elan @pixellight @laser ah, because your gender is country-specific.
What I want to know is what you've got to do to wind up ` IS_A_COUPLE`.
@zbrown @elan @pixellight I'm wondering if Brangelina etc would trip it
@laser looks like privacy violation as a service.
@laser Ignoring the ickiness of it, thankfully it's also just not going to work because these people don't appear to understand how names work.
Took me literally *one* test to find a name it fails on. "Janne" is a first name common in places w/Germanic languages, which is mostly male some places, female some places, and either elsewhere. The API claims it's male w/100% certainty, which isn't true *anywhere* thanks to immigration.
They have way too little data to even know how wrong they are.
@adenner @laser I think their idea is things like stripping down registration forms, or applying demographics for ad targeting purposes and the like.
But of course, the former falls down not just based on misclassifying names but people who prefers a form of address that doesn't match their assigned sex, and the latter is just just creepy.
I'm really curious about what kind of attempts they've made to validate product fit here, because it seems so obvious that it's a bad idea.
@laser I’m a dude named Cary who can’t get his name spelled right even when it’s right there in the header of an email. COME AT ME, SaaS!!!!!
@RachaelAva1024@laser This isn't what I had in mind when I thought "Gender API" and honestly, I'm disappointed. 
@RachaelAva1024 the REST API doesn't even support PATCH requests smh
@laser I was scrolling along and saw "genderize your data" and assumed it was a joke a transgender person was making, the way we do, but what the hell, it is not!
@laser apparently, it's computer automation of assuming people's genders
@laser i’ve checked their site and it’s even worse there:
“Gender-API helps you to find out whether a first name is more likely be used by males or females.”
imagine dehumanizing your customers
@laser SaaS (Stereotyping as a Service)
@laser MEAT GENDERIZER.
@ceralor@blimps.xyz@laser we live in hell already actually
Qual é o seu gênero?
@laser not even good, extremely french names have apparently turned out German according to it lol
@dalite@bark.lgbt@laser This is totally not ok what the hell
>Entire over decade controversy about assuming people's gender being bad
>"Let a robot guess and we'll pretend it's fact"
@laser an undergrad trained a super basic ai model on names and ages and genders and then got a ton of funding because VCs are idiots that will fund anything, probably
@lunch I don't think it's even a proper classifier, it's probably just a wrapper around the PHP extension so they didn't do the work themselves https://github.com/cstuder/genderReader/tree/master/gender.c
@laser i've looked into this here's what it actually does:
- given a first name, looks it up in a table and give you male/female with a confidence %
- given an email, uses shitty heuristics to extract a first name, and then does the above
that's it!
@laser like it's not even Machine Learning or whatever buzzword would have gotten them more money it's just a big table of (publicly available elsewhere) data that you pay them to be able to look stuff up in
@a11ce apparently there's a PHP extension that does this so quite a few people who've responded to me just think it's wrapping that so it's not even original work
@laser I guess "Clockr" was taken.
@laser gonna headcanon that genderized data means like
"Finally! CSV files, for Men 
" or "Introducing: 
Relational databases, for Her 
"
@laser AYGaaS: assuming your gender as a service
@laser but like seriously, even if you assume that everybody is stereotypically binary cis this is still ripe for making mistakes and guessing wrong
i remember seeing this some months ago and at least when i looked into it (I'm assuming it hasn't really changed) and like it makes the assumption all names have a gender, it has a confidence score but that doesn't really do anything to disambiguate gender neutral names (and i don't trust that those who'd use this API are gonna use the confidence value properly)
i think it has a region or language field in the api, but that still doesn't help if someone from say italy moves to america and picks english as their language of choice, and i have doubts it'll handle names from minority languages properly
at the end of the day, if you need a gender for whatever reason just ask, don't try to silently guess it