My favourite bit is actually that the paper starts with a HUGE disclaimer saying they're aware that inferring gender from first names is inaccurate and causes harm especially to marginalised groups, AND THEN THEY DO IT ANYWAY.
Shockingly, the software doesn't gender "Cyrridven" or "Kerewin" correctly. Can you IMAGINE?
I guess professionally my gender is andy, according to gender-guesser:
>>> import gender_guesser.detector as gender
>>> d = gender.Detector()
>>> d.get_gender("Wen")
andy
re: nudity, DeepNude
On a more serious note, this software doesn't work for pictures of me wearing anything more than swimwear, so I'm gonna continue to be more worried about the abuse of face-swapping networks rather than DeepNude networks.
@RaeHaskell@twitter.com @fancytypes@twitter.com Could you explain why this happens? Is this expected, or an implementation bug?
There's something wonky about closed type families!
If both arguments to 'Min' contain type variables, only the first rule is ever tried. The family below gives 'Min n n ~ n' but if we move the first rule down, it no longer computes!
https://gist.github.com/wenkokke/7ffe68c9c2dc06679d9e6f63d2bcad17
proctoring software
Cop shit has no place in the classroom and fucks over poor, Black, and disabled students.
RT @MathProfPeter@twitter.com
In my in-person sessions with students today (for m hybrid courses), they were telling me about their experience using proctoring software and lockdown browsers for OTHER courses (I could never...). Here are some of the things that they just said casually:
🐦🔗: https://twitter.com/MathProfPeter/status/1364072924550991873
Heya @kosmikus@twitter.com! I wrote some lhs2TeX directives to format Linear Haskell's arrow as %1 ⊸ based on lambda.fmt. Unfortunately, I think the linear arrow syntax is too complex to conditionally hide the %1 using JUST lhs2TeX. Any suggestions? :)
https://gist.github.com/wenkokke/97b9055aae4c05842a16713a722839ff