This article is very uninformative. What’s the evidence this “safety label” is being used in an improper way and not for example Russian state propaganda or some of the extremely violent war footage?
If Twitter doesn't want people to draw the wrong conclusions, maybe they can release the whole actual algorithm? Not just some weird category preprocessing whatever thing?
As it stands, they throw this out there, their PR email responds with poop, Elon is in some space spreading misinformation/having no clue what he actually released.
"Where is the war footage?!" is one of the main talking points among far-right disinfo accounts to dowplay the severity of Russia's invasion.
Selectively down-ranking war footage from Ukraine is suspicious when they don't do the same for war footage from other parts of the world.
I actually think it's a good thing that twitter doesn't censor or down-rank war footage from the Middle East showing the results of American and Israeli bombings.
However that just makes it more suspicious that they are down-ranking footage that show the results of Russia's invasion of Ukraine.
Just in case anyone is curious, it is on Telegram and various western commentators' Patreon accounts. Probably unsurprising, but you will see people die if you watch it.
Programmers being bad at naming things isn’t very damning. Add to that this guy made a few 1000 enemies with inside info with his hamfisted management, it would be leaked to the press 5 minutes after he orders it.
Btw this class seems to be for spaces not tweets and a quick search for Ukraine in
TweetSafetyLabels in the same package turns up no results, really shoddy writing.
I noticed that too, but it's likely we are not seeing the whole code base. "the algorithm", despite being possibly at the heart of many "interesting" things, is likely just one out of many other "modules" of the Twitter platform.
Also while it was just rumours so far, various well known Tweeter tweeting about this war have been complaining since last November that their tweets had suddenly much less engagements. This is just too much of a coincidence.
> Programmers being bad at naming things isn’t very damning
You’re right. But Musk was hosting Twitter polls, posting memes and proposing territorial borders and conflict resolution that seemed to be much closer to the Russian view than the Ukrainian view.
So you’re right that more info is needed, but Musk inserted himself into this topic in a way that raised a few eyebrows.
If Twitter doesn't want people to draw the wrong conclusions, maybe they can release the whole actual algorithm? Not just some weird category preprocessing whatever thing?
As it stands, they throw this out there, their PR email responds with poop, Elon is in some space spreading misinformation/having no clue what he actually released.