anonymised people of the global south

Oh, to live in the present where infinite potential exists! We all have a hand in the wake that will reverberate

it’s so wild that trumps executive order has that bit about sex being assigned at conception.

I assume this is the anti-abortion position being encoded into the anti-trans legislation. no one is testing what gametes their embryo has— such a thing isn’t possible (embryos don’t produce them yet; there are primordial germ cells, but they aren’t differentiated as sperm or egg cells).

it just goes to show that gender/sex policing is never based on any science— there is no "scientific" definition to what “female sex” or "male sex" is. It doesn’t exist as a material reality. It is a legal construct that is assigned with the first legal documentation assigned to a subject—the birth certificate.

and that isn’t based on chromosomes or gametes. it’s a decision based on whether a doctor thinks the infant will be mostly likely to be capable of penetrative heterosexual sex as “female” or “male.” because penetrative heterosexual sex is the point, the determination is based on whether the clitoris is long enough to penetrate (in which case “it’s a boy”) and if not whether there is a vaginal canal deep enough for penetration (in which case “it’s a girl”) and if neither or both are true the child’s body is forcibly surgically altered.

they can't define sex "biologically" because it is not a "biological" reality. it will be determined, as always, punitively, on a case by case basis on varying traits and legal documentation, to enforce patriarchal economic systems that depend on misogyny and transmisogyny.

one of the most telling parts from the executive order, to me, was this (emphasis mine):

this bit gives the game away with how important the institution and enforcement of sex assignment is to capitalism, the united states economy, the carceral/legal system, and the ideological buy-in to us-american hegemony.

American media and the likes of ADL are insulting our intelligence by claiming this is not a nazi salute.

it would be awesome if mushrooms were high protein and/or high calorie because they are so tasty but sadly they're basically just water... like i should be able to use them as my main protein source

i love checking the weather. at least we have the weather

it's hard work maintaining the levels of ignorance about music and cinema that i exhibit 👍🏽

i went to check this "angels" by this "robbie williams" bc with famous songs it's v common that ive heard them multiple times and even liked, just don't remember what they're called or who sang them, but this one i have literally never heard in my life

"don't make fun of people for not knowing something" ya but you understand there's a difference between actively making racist & dismissive assumptions about chinese people and like, not knowing exactly how chinese housing rentals work or whatever. right.

19th century ethnologists were expressing their horizon-expanding curiosity in an enlightening process of cultural exchange

To understand what's going on here, know these things:

It's absurd and horrifying at the same time.

This is neither absurd nor horrifying, and two statements here are false (4 and 6).

1.8m of the 3m page requests were to robots.txt. If they weren't respecting robots.txt, then the site admin would see many, many more overall requests, and a significantly lower fraction of them would be to robots.txt.

The email explicitly says they have several billion websites, not just pages. In this case, it looks like they're all on different subdomains. There isn't a standard way to have robots.txt apply to all subdomains, which is why the person said robots.txt isn't a good solution.

Everyone doing LLM training does not just dump all the data directly into the training pipeline. Data processing happens before that step, but first they need to collect data to even process; that's what the crawler is. Part of the data processing step, after the crawling step, involves deduplication and quality filtering. It's misleading to say this is data they're training on; this is data going into the pool of stuff that might be trained on. That raw crawl has untold billions of pages no doubt filled with higher and lower quality garbage that will get filtered out before training.

The solution they're looking for is just to contact someone to have them blacklist either the whole domain or the IP address, since it's apparently all running from the same IP. With large crawlers, shit like this happens all the time—they even mention the same thing happened with Amazon.

This honestly isn't very exciting.

You can literally go to the relevant robots.txt file and examine it, all lines are commented out aside from the ones that tell *all* bots to not go to the /archive subsection of the site, which is presumably there to test if the bots are listening to instructions. Everything else is *allowed*.

OpenAI's web crawler downloads the robots.txt file, and handles it *correctly*.

The failure on OpenAI's side is by not having limits on how deep to go.

TL;DR: OpenAI's spider is not ignoring the sign that tells it to not go there, it reads it, correctly understands that it can go there, goes there and enters a web so complicated it gets lost and is unable to leave.

these violent delights kick ass. anyone know how they end?

Angelic Page (Urza's Saga) - Rebecca Guay

More cards with art by Rebecca Guay on Scryfall

Rhyton

Terracotta Vase in the Form of a Lobster Claw ca. 460 BC.

Greek