During the brazilian elections we learned one obvious but important tactic to defeat the far right on the Internet and I think everyone should know it:
IGNORE THEM! DON'T ENGAGE WITH THEIR CONTENT! DON'T MAKE CONTENT REACTING TO THEM! DON'T HATE QRT! LET THEM ALONE IN THEIR BUBBLE!
Josh Turek of Iowa, Democrat, comes out with one of the worst anti-trans ads yet.
Not only is it an anti-sports ad, it proposes a ban on federal funding for gender affirming care. That would hit all medicaid and medicare patients, and any ACA federally-subsidized plans as well.
Horrific.
โBefore July 2025, one of the food bankโs locations served between 20 and 40 families per day, he said. Now, more than 260 families come in daily for helpโฆโ
About 2 million kids have lost SNAP since deep cuts were enacted last year. The ripple effects: Kids losing SNAP also lose automatic enrollment for free meals at school. And as more kids lose SNAP, their schools may be forced to drop Community Eligibility.
It's important to know how LLMs work: they predict how some text ends. That's it.
That's why the early ones were literally "enter some text, and it'll finish it."
The "magic" is that the text they see isn't just your input. It's surrounded by context, and filtered.
Since the exact expression is unlikely to be in its model, it reads like:
"The user is asking if SOME_NUMBER [...] since this" - completing from there, any response is likely.
Newer models try to "fix" this with more data, "peer review" by other models, etc, but the fundamentals are the same.
So, despite advances that make them pretty good at faking it, whenever you ask them anything, what you're really asking is "statistically, what's the most likely response to a question like this?"
It's important to know how LLMs work: they predict how some text ends. That's it.
That's why the early ones were literally "enter some text, and it'll finish it."
The "magic" is that the text they see isn't just your input. It's surrounded by context, and filtered.
Stage 2 input: "[...] Its response is: 333333.33333333"
Feeding this back in again, it continues matching the pattern:
"Its response is: 333333.33333333 - since this contains a decimal, the answer is No. I should reply: No, 1 million is not divisible by 3."
This is all well and good, but the final response still came from just statistical analysis of language. Earlier models were just as likely to finish it with "since this doesn't contain a decimal" or "since this contains a decimal, the answer is Yes". Those are all likely endings.
It's important to know how LLMs work: they predict how some text ends. That's it.
That's why the early ones were literally "enter some text, and it'll finish it."
The "magic" is that the text they see isn't just your input. It's surrounded by context, and filtered.
Input after stage 1: "The user is asking if 1000000 is divisible by 3. This is a math expression, so I should ask the Calculator tool. Its response is: <calc:1000000/3>"
The system sees that tag and runs "calc 1000000/3", replacing the tag with its result.
It's important to know how LLMs work: they predict how some text ends. That's it.
That's why the early ones were literally "enter some text, and it'll finish it."
The "magic" is that the text they see isn't just your input. It's surrounded by context, and filtered.
You enter "hi" and it sees sth like:
"You are an advanced chatbot, [various rules and character description], the user says: hi; you reply: "
That's why injection can work. It's still just using the structural patterns of language to determine the most likely followup, no understanding needed.
For two years, student-led protests have raged against Aleksandar Vuฤiฤ and his corrupt government in Serbia.
Today, in parliamentary elections, he was toppled.
One of the most sustained protest movements in recent European history has succeeded.
Resistance works.
Your monthly reminder that if you're using ChatGPT to do your science, you're failing your colleagues, community, and all of us.
Here, ChatGPT makes up data and presents it as the real thing โ then apologizes when confronted.
This so much. I'm always like "I'm willing to follow the rules but it would be a lot easier if I understood the reasoning behind them?" and then get labeled as refusing to follow the rules
Some rules exist because people just kept following them without knowing why and they actually no longer apply.
Classic story: "why always cover your roast while it cools?" "That's what you did" "that was to keep cats away! You don't have cats!"
Or the experiment where all monkeys are punished when one goes for the treat atop the tower.
They learn not to do it. Newcomer arrives. They stop him.
Eventually all that were punished have been replaced. They still stop newcomers from climbing. None know why. Not even a treat up there anymore.
This so much. I'm always like "I'm willing to follow the rules but it would be a lot easier if I understood the reasoning behind them?" and then get labeled as refusing to follow the rules
"because I said so" is not a valid answer
You might be tempted to use it when you feel the real answer is too difficult to explain to a kid, but just... say that. "I'll explain later, just trust me."
"I don't know" is also valid, but I strongly encourage you to find out.
NEW: Epic Games has delayed its upcoming (not yet announced) Star Wars + Marvel shooter as it pivots to complete AI game development.
"No more hand-written code," a director told staff. All art "generated with full AI workflows."
Full scoop: www.bloomberg.com/news/article...
Boy, phones sure are great. I love being forced to update and then suddenly my device is "not compatible" with an app that's worked for years. But at least now some storage space is going to built-in parental control features I have no use for.
Hi. I, a trans woman, am the Executive Producer of this show. I wrote the question. He called trans girls men. He has now said it in three different places in the past week.
God I am so fucking angry
Queer folks, tell your stories. Mainstream, independent, comics, zines, games, shorts, novels, film, whatever your medium whatever your genre
There are increasingly fewer platforms for them, so it is more important than ever to make sure they get out into the world
Am I a sex worker?
No.
Do I use the services of sex workers?
Also no.
Do I need someone to explain to me why platforms like OF, which allow SWs to control their own services w/o an employer or โmanagerโ are constantly attacked but you can sign into Pornhub w Google?
Yeah man. Yeah I do.